.

T H E ORIGINAL D E S I G N ANE
ULTIMATE DESTINY O F THE
WORLD WIDE WEB
â„¢SSSSSSBSE "[An] important account
••••••••••••••fe of how, when, where,

ihhSESSSSSSSS! " a n d w h y
[ B e r n e r s
- L e e
i
cooked up the w e b â€"
•5Sh5SS™™^*™WP well worth reading."
â€" N E W YORK TIMES

" ^ ^ • • • • • • P * BOOK REVIEW




F o r e w o r d by M I C H A E L D E R T O U Z O S , Director of MIT Laboratory for Computer Scienct




TIM BERNERS-LEE
W I T H M A R K F I S C H E T T I
W pa v i n g
*W.ebt h


THE ORIGINAL DESIGN
and ULTIMATE DESTINY

of the W O R L D WIDE W E B

BY ITS INVENTOR



TIM BERNERS-LEE
with MARK FISCHETTI




m HarperBusiness
An Imprint of HirperCoQinsPttblisbers
 C o n t e n t s




A C K N 0 V/ L E C G E NTS v




FOREWORD BY MICHAEL DERTOJ20S vii



1 Enquire W i t h i n upen Everything i


To Nancy 2 T a n g l e s , L i n k s , and W e b s 7

3 ¡nFo.cern.ch 25

4 Protocols: Simple Rules For

Global Systems 35

WEAVING THE WEB: Tlie Original Design and Ultimate Destiny of the World Wide Web 5 Going Global 53
by Its Inventor. Copyright ' 1999, 2000 by T i m Bemcrs-Lee. All rights reserved.
Printed in the United States of America. No part of this book may be used or
6 Browsing 57
reproduced in any manner whatsoever without written permission except in the
case of brief quotations embodied in critical articles and reviews. For information
address HarperCollins Publishers Inc., 10 East 53rd Street, New York, N Y 10022. 7 Changes 75

HarperCollins books may be purchased for educational, business, or sales pro- fi Consortium 91
motional use. For information please write: Special Markets Department,
HarperCollins Publishers Inc., 10 East 5 3 rJ
Street, New York, N Y 10022.
? Competition and C o n s e n s u s 103
HarperCollins Web Site: www.harpercollins.com
10 Web oF P e o p l e 123
First paperback edition published 2000.

Designed by Laura Lindgren and Celia Fuller 11 Privacy 143

Library of Congress Cataloging-in-Publication Data 12 Mind to Mind 157
Berners-Lee, T i m .
Weaving the Web : the original design of the World Wide Web by its
inventor / T i m Berners-Lee with Mark Fischetti. â€"1st ed. 13 M a c h i n e s and t h e Web 177
p. cm.
ISBN 0 - 0 6 - 2 5 1 5 8 7 - X (paper) 14 Weaving the Web 199
1. World Wide Web â€"History. 2. Berners-Lee, T i m . I. Fischetti, Mark.
APPENDIX:
II. Title.
TK5105.888.B46 2000 INFORMATION MANAGEMENT: A PROPOSAL 211

025.04-dc21 00-039593 GLOSSARY 231

08 «WRRD 10 y 8 7 ft 5 IMC EX 241
 A c k n o w l e d g m e n t s




A book is quite a project. I had thought about one f r o m time to
time, but did not take it on u n t i l Michael Dertouzos introduced
me to M a r k Fischetti as someone w h o , unlike me, could actually
make it happen w i t h o u t stopping everything for a year. A n d so
began the telling of the story, past, present, and future. Without
M a r k this book w o u l d never have been more than an idea and
some bits of unordered web pages. I owe great thanks to M a r k
for applying his ability to find the thread running through m y
incoherent ramblings and then a way to express it simply.
M a r k and I together owe thanks to everyone else involved i n
this process: to Michael for the idea of doing it and the encour-
agement, to Ike Williams for organizing it, and to Liz Perle at
Harper San Francisco for her excruciating honesty and insistence
that the book be what it could be. W i l l i a m Patrick played a great
role i n that step, helping us get it to a f o r m w i t h w h i c h we were
all happier. We all have to thank Lisa Zuniga and the production
team for turning the bits into a book. If you are reading this on
paper, then the miracle of coordination must have been pulled off
despite all m y missed deadlines.
Many of these people mentioned have suffered the shock of
meeting m y stubbornness at wanting to call the shots over working

v
 a c k n o w l e d g m e n t s


methods and ways of transferring data. I apologize . . . this time: F o r e w o r d
Next time, w e ' l l do it all online!
The book owes its existence indirectly to everyone w h o has
been involved i n making the dream of the Web come true. One of
the compromises that is part of a book is that some occasions and
activities t u r n out to be appropriate for showing what life was
like and w h a t the principles behind it were. Others, w h i l e just as
important, don't turn up as examples i n the narrative. So the
index of the book doesn't serve as a hall of fame, as plenty of
people have necessarily been left out or, perhaps even more
strangely, it was only practical to describe one particular part of
their many contributions. A l l the consortium team (W3T), pres-
ent and alumni (listed on the w w w . w 3 . o r g site), are priceless
peopleâ€"working w i t h them is great. Weaving the Web is a unique story about a unique innovation, by
I w o u l d like to thank permanently, irrespective of this book, a unique inventor.
everyone w h o has taken time out to move the Web onward for A m i d the barrage of information about the World Wide Web,
the common good. For everyone w h o has helped, there have also one story stands out â€"that of the creation and ongoing evolution
been the managers and family w h o actively or passively provided of this incredible new thing that is surging to encompass the
encouragement. For me, the managers were Peggie Rimmer and w o r l d and become an important and permanent part of our his-
Mike Sendall at CERN, whose w i s d o m and support have been tory. This story is unique because it is w r i t t e n by T i m Berners-
very special to me. Lee, w h o created the Web and is now steering it along exciting
To thank my immediate family here w o u l d suggest I were future directions. No one else can claim that. A n d no one else
only thanking them for helping w i t h the book, and for putting up can w r i t e this â€"the true story of the Web.
w i t h my strange behavior during book crises. The support you Tim's innovation is also unique. It has already provided us
three have given me is more than t h a t â€" i t is a sense of perspec- w i t h a gigantic Information Marketplace, where individuals and
tive and reality and f u n that underlies everything we do, of organizations buy, sell, and freely exchange information and infor-
w h i c h the Web and this has been one, though a notable, part. mation services among one another. The press, radio, and televi-
T i m Berners-Lee sion never got close; all they can do is spray the same information
Cambridge, Massachusetts out f r o m one source toward many destinations. Nor can the letter
or the telephone approach the Web's power, because even though
those media enable one-on-one exchanges, they are slow and
devoid of the computer's ability to display, search, automate, and
mediate. Remarkably â€"compared w i t h Gutenberg's press, Bell's
telephone, and Marconi's radio â€"and well before reaching its

VI VLL
F o r e w o r d F o r e w o r d


ultimate form, Berners-Lee's Web has already established its and their social impact. Many people in the w o r l d believe that
uniqueness. technology is dehumanizing us. At LCS, we believe that technol-
Thousands of computer scientists had been staring for two ogy is an inseparable child of humanity and that for true progress
decades at the same two things â€"hypertext and computer net- to occur, the two must w a l k hand i n hand, w i t h neither one act-
works. But only T i m conceived of how to put those two elements ing as servant to the other. I thought it w o u l d be important and
together to create the Web. What k i n d of different thinking led interesting for the w o r l d to hear f r o m the people w h o create our
h i m to do that? No doubt the same thinking I see driving h i m future rather than f r o m some sideline futurologists â€"especially
today as he and the World Wide Web Consortium team that he when those innovators are w i l l i n g to expose the technical forces
directs strive to define tomorrow's Web. While the rest of the and societal dreams that drove them to their creations. T i m has
w o r l d is happily mouthing the mantra of electronic commerce, risen to this challenge admirably, exposing his deep beliefs about
he is thinking of the Web as a medium that w o u l d codify, i n its how the Web could evolve and shape our society i n ways that are
gigantic distributed information links, human knowledge and fresh and differ markedly f r o m the common w i s d o m .
understanding. In Weaving the Web, T i m Berners-Lee goes beyond laying out
W h e n I first met T i m , I was surprised by another unique trait the compelling story of the Web: He opens a rare w i n d o w into
of his. As technologists and entrepreneurs were launching or the way a unique person invents and nurtures a unique approach
merging companies to exploit the Web, they seemed fixated on that alters the course of humanity.
one question: ' H o w can I make the Web mine ?'" Meanwhile, Tim
1
Michael L. Dertouzos
was asking, "How can I make the Web yours?" As he and I began
planning his arrival at the M I T Laboratory for Computer Science Michael L. Dertouzos is the director of the M I T Labora-
and the launching of the World Wide Web Consortium, his con- tory for Computer Science and the author of the book
sistent aim was to ensure that the Web would move forward, WJiat Will Be.
flourish, and remain whole, despite the yanks and pulls of all the
companies that seemed bent on controlling it. Six years later,
Tim's compass is pointed i n exactly the same direction. He has
repeatedly said no to al! kinds of seductive opportunities if they
threatened, i n the least, the Web's independence and wholeness,
and he remains altruistic and steadfast to his dream. I am con-
vinced that he does sc not only f r o m a desire to ensure the Web's
future, but also f r o m a wellspring of human decency that I find
even more impressive than his technical prowess.
When I first suggested to T i m that he w r i t e this book, and
having just finished one myself, I was envisioning a series of
books f r o m the M I T Laboratory for Computer Science (LCS) in
w h i c h we w o u l d discuss i n everyday language our innovations
VLLL
 C H A P T E R




E n q u i r e W i t h i n

u p o n E v e r y t h i n g




W h e n I first began tinkering w i t h a software program that even-
tually gave rise to the idea of the World Wide Web, I named it
Enquire, short for Enquire Within upon Everything, a musty old
book of Victorian advice I noticed as a child i n m y parents' house
outside London. W i t h its title suggestive of magic, the book
served as a portal to a w o r l d of information, everything f r o m how
to remove clothing stains to tips on investing money. Not a per-
fect analogy for the Web, but a primitive starting point.
What that first bit of Enquire code led me to was something
much larger, a vision encompassing the decentralized, organic
growth of ideas, technology, and society. The vision I have for the
Web is about anything being potentially connected w i t h anything.
It is a vision that provides us w i t h new freedom, and allows us to
grow faster than we ever could w h e n we were fettered by the
 w e a v i n g t h e w e b e n q u i r e w i t h i n u p n n e v e r y t h i n g


hierarchical classification systems into w h i c h we bound our- ing the Web at this deeper level w i l l people ever truly grasp what
selves. It leaves the entirety of our previous ways of w o r k i n g as its f u l l potential can be.
just one tool among many. It leaves our previous fears for the Journalists have always asked me what the crucial idea was,
future as one set among many. A n d it brings the workings of soci- or what the singular event was, that allowed the Web to exist one
ety closer to the workings of our minds. day w h e n it hadn't the day before. They are frustrated w h e n I
Unlike Enquire Within upon Everything, the Web that I have tell them there was no "Eureka!" moment. It was not like the leg-
tried to foster is not merely a vein of information to be mined, endary apple falling on Newton's head to demonstrate the con-
nor is it just a reference or research tool. Despite the fact that the cept of gravity. Inventing the World Wide Web involved m y
ubiquitous www and .com now fuel electronic commerce and stock growing realization that there was a power i n arranging ideas i n
markets all over the world, this is a large, but just one, part of the an unconstrained, weblike way. A n d that awareness came to me
Web. Buying books from Amazon.com and stocks from E-trade is through precisely that k i n d of process. The Web arose as the
not all there is to the Web. Neither is the Web some idealized space answer to an open challenge, through the swirling together of
where we must remove our shoes, eat only fallen fruit, and eschew influences, ideas, and realizations f r o m many sides, until, by the
commercialization. wondrous offices of the human mind, a new concept jelled. It
The irony is that i n all its various guises â€"commerce, research, was a process of accretion, not the linear solving of one well-
and surfingâ€"the Web is already so much a part of our lives that defined problem after another.
familiarity has clouded our perception of the Web itself. To I am the son of mathematicians. M y mother and father were
understand the Web in the broadest and deepest sense, to fully part of the team that programmed the world's first commercial,
partake of the vision that I and m y colleagues share, one must stored-program computer, the Manchester University "Mark I , "
understand how the Web came to be. w h i c h was sold by Ferranti L t d . i n the early 1950s. They were
The story of how the Web was created has been told i n various f u l l of excitement over the idea that, i n principle, a person could
books and magazines. Many accounts I've read have been distorted program a computer to do most anything. They also knew, how-
or just plain wrong. The Web resulted from many influences on ever, that computers were good at logical organizing and process-
my mind, half-formed thoughts, disparate conversations, and seem- ing, but not random associations. A computer typically keeps
ingly disconnected experiments. I pieced it together as I pursued information i n rigid hierarchies and matrices, whereas the
my regular w o r k and personal life. I articulated the vision, wrote human m i n d has the special ability to link random bits of data.
the first Web programs, and came up w i t h the now pervasive W h e n I smell coffee, strong and stale, I may find myself again i n
acronyms URL [then UDI), HTTP, H T M L , and, of course, World a small room over a corner coffeehouse i n Oxford. M y brain
Wide Web. But many other people, most of them unknown, con- makes a link, and instantly transports me there.
tributed essential ingredients, i n much the same almost random One day w h e n I came home f r o m high school, I found m y
fashion. A group of individuals holding a common dream and father w o r k i n g on a speech for Basil de Ferranti. He was reading
working together at a distance brought about a great change. books on the brain, looking for clues about how to make a com-
M y telling of the real story w i l l show how the Web's evolu- puter intuitive, able to complete connections as the brain did. We
tion and its essence are inextricably linked. Only by understand- discussed the point; then m y father went on to his speech and I
2 3
 w e a v i n g t h e w e b e n q u i r e w i t h i n u p o n e v e r y t h i n g


went on to m y homework. But the idea stayed w i t h me that com- Computers might not find the solutions to our problems, but
puters could become much more powerful if they could be pro- they would be able to do the b u l k of the legwork required, assist-
grammed to l i n k otherwise unconnected information. • u our human minds i n intuitively finding ways through the
m


This challenge stayed on m y m i n d throughout m y studies at maze. The added excitement was that computers also could follow
Queen's College at Oxford University, where I graduated i n 1976 and analyze the tentative connective relationships that defined
w i t h a degree i n physics. It remained i n the background w h e n I much of our society's workings, unveiling entirely new ways to
built m y o w n computer w i t h an early microprocessor, an old tele- see our w o r l d . A system able to do that w o u l d be a fantastic thing
vision, and a soldering iron, as w e l l as during the few years I for managers, for social scientists, and, ultimately, for everyone.
spent as a software engineer w i t h Plessey Telecommunications Unbeknownst to me at that early stage i n m y t h i n k i n g , sev-
and w i t h D . G . Nash Ltd. eral people had hit upon similar concepts, w h i c h were never
Then, i n 1980, I took a brief software consulting job w i t h implemented. Vannevar Bush, onetime dean of engineering at
CERN, 1
the famous European Particle Physics Laboratory i n MIT, became head of the U.S. Office of Scientific Research and
Geneva. That's where I wrote Enquire, m y first weblike program. Development during W o r l d War I I and oversaw development of
I wrote it i n m y spare time and for m y personal use, and for no the first atomic bomb. I n a 1945 article i n the Atlantic Monthly
loftier reason than to help me remember the connections among titled "As We M a y T h i n k , " he w r o t e about a photo-electro-
the various people, computers, and projects at the lab. Still, the mechanical machine called the Memex, w h i c h could, by a
larger vision had taken firm root i n m y consciousness. process of binary coding, photocells, and instant photography,
Suppose all the information stored on computers everywhere were make and follow cross-references among m i c r o f i l m documents.
linked I thought. Suppose I could program my computer to create a Ted Nelson, a professional visionary, wrote i n 1965 of "Liter-
space in which anything could be linked to anything. A l l the bits of ary Machines," computers that w o u l d enable people to w r i t e and
information i n every computer at CERN, and on the planet, publish i n a new, nonlinear format, w h i c h he called hypertext.
w o u l d be available to me and to anyone else. There w o u l d be a Hypertext was "nonsequential" text, i n w h i c h a reader was not
single, global information space. constrained to read i n any particular order, but could follow links
Once a bit of information i n that space was labeled w i t h an and delve into the original document f r o m a short quotation. Ted
address, I could tell m y computer to get i t . By being able to refer- described a futuristic project, Xanadu, i n w h i c h all the world's
ence anything w i t h equal ease, a computer could represent asso- information could be published i n hypertext. For example, if y o u
ciations between things that might seem unrelated but somehow were reading this book i n hypertext, you w o u l d be able to follow
did, i n fact, share a relationship. A web of information w o u l d a link f r o m m y reference to Xanadu to further details of that pro-
form. ject. I n Ted's vision, every quotation w o u l d have been a l i n k back
to its source, allowing original authors to be compensated by a
very small amount each time the quotation was read. He had the
1 The name C E R N derives from the name of the international council (Conseil
dream of a Utopian society i n w h i c h all information could be
Européen pour la Recherche Nucléaire], w h i c h originally started the lab. The
council no longer exists, and "Nuclear" no longer describes the physics done
shared among people w h o communicated as equals. He struggled
there, so while the name C E R N has stuck, it is not regarded as an acronym. for years to find funding for his project, but success eluded h i m .
4 5
w e a v i n g t h e w e b


Doug Engelbart, a researcher at Stanford University, demon-
strated a collaborative workspace called NLS |oN Line System) i n
the 1960s. Doug's vision was for people to use hypertext as a tool
for group w o r k . I n order to help himself steer his computer's cur-
sor across the screen and select hypertext links w i t h ease, Doug
invented a wooden block w i t h sensors and a ball underneath, T a n q I e s,
and called it a mouse. I n a now-famous video, w h i c h I didn't see
until 1994, Doug demonstrated using electronic mail and hyper- L i n k s , and W e b s
text links w i t h great agility w i t h his homemade mouse i n his
right hand and a hve-key piano-chord keyboard i n his left hand.
The idea was that a person could interface w i t h the machine i n a
very close, natural way. Unfortunately, just like Bush and Nelson,
Doug w as too far ahead of his time. The personal computer revo-
r




lution, w h i c h w o u l d make Engelbart's "mouse" as familiar as the
pencil, w o u l d not come along for another fifteen years. W i t h that
revolution, the idea of hypertext w o u l d percolate into software
design.
Of course, the next great development i n the quest for global
connectivity was the Internet, a general communications infra-
structure that links computers together, on top of w h i c h the Web
The research center for particle physics known as CERN straddles
rides. The advances by Donald Davis, by Paul Barran, and by
the French-Swiss border near the city of Geneva. Nestled under
Vint Cerf, Bob Kahn, and colleagues had already happened i n the
1970s, but were only just becoming pervasive. the limestone escarpments of the j u r a mountains, ten minutes
from the ski slopes, w i t h Lac Leman below and M o n t Blanc
I happened to come along w i t h time, and the right interest
and inclination, after hypertext and the Internet had come of age. above, it offered unique research opportunities, and the area

The task left to me was to marry them together. offered a very pleasant place to live.
Engineers and scientists arrived at CERN f r o m all over the
w o r l d to investigate the most fundamental properties of matter.
Using enormous machines, they w o u l d accelerate tiny nuclear
particles through a series of tubes that, though only a few inches
wide, ran for several kilometers w i t h i n a m a m m o t h circular
underground tunnel. Researchers w o u l d rev up the particles to
extremely high energies, then allow them to collide. For an
unimaginably brief instant, new particles might be made, then

6
 t a n q ' e s , l i n k i , , a n d w e b * ,



lost. The trick was to record the high-energy debris f r o m the cat- control room, to actually program a computer system. Kevin and
aclysm as it careered into one of two detectors inside the tunnel, I would soon j o i n a team of people w h o w o u l d ultimately bring
each the size of a house, jammed f u l l of electronics. about the demise of that control room. Alas, the racks of glowing
Research on this scale was so expensive that it had to involve electronics w o u l d be slowly dismantled and replaced by a boring
collaborations among many nations. Visiting scientists w o u l d r u n oval of computer consoles, r u n by m u c h more powerful software.
their experiments at CERN, then go back to their home institu- The big challenge for contract programmers was to t r y to
tions to study their data. Though it was a central facility, CERN understand the systems, both human and computer, that ran this
was really an extended community of people w h o had relatively fantastic playground. M u c h of the crucial information existed
little common authority. The scientists brought a w i d e variety of only i n people's heads. We learned the most i n conversations at
computers, software, and procedures w i t h them, and although coffee at tables strategically placed at the intersection of two cor-
they came f r o m different cultures and spoke different languages, ridors. I w o u l d be introduced to people plucked out of the flow of
they managed to find a way to w o r k together because of their unknown faces, and I w o u l d have to remember w h o they were
shared interest i n particle physics and their desire to see a huge and w h i c h piece of equipment or software they had designed.
project succeed. It was a tremendously creative environment. The weblike structure of CERN made the job even harder. Of the
I n 1980, CERN was i n the process of replacing the control ten thousand people in the CERN phone book, only five thousand
system for two of its particle accelerators. The w o r k was getting or so were at CERN at any given time, and only three thousand
behind, and CERN needed help. I had, by chance, been consult- or so were actually salaried staff. Many of the others had a desk,
ing elsewhere i n Switzerland w h e n m y friend and colleague and visited f r o m their home institutions only every now and
Kevin Rogers called f r o m England to suggest we apply. again.
Upon our arrival to be interviewed, Kevin and I were given a To house contractors w h o suddenly arrived i n a panic to help
tour, and soon found ourselves on a catwalk, looking out and advance some project or other, management had erected portable
over what looked like a huge, chaotic factory floor. This vast cabins on the top of a grassy h i l l on the grounds. Groups of us
experimental hall was filled w i t h smaller experiments, obscured would discuss our ideas at lunch overlooking the Swiss vineyards,
by the concrete-block walls between them, hastily built to cut or as we walked down the long flight of concrete steps f r o m the
down radiation. Continuing along the catwalk, we came to the hill to the experiment hall and terminal room to do the program-
control room. Inside were racks and racks of computing hard- ming. I filled in the odd moments w h e n I wasn't officially work-
ware, w i t h no lighting except for the glow of the many indicator ing on the Proton Synchrotron Booster by tinkering w i t h m y play
lamps and dials. It was an electronic engineer's paradise, w i t h program, the one I called Enquire. Once I had a rough version, I
columns of oscilloscopes and power supplies and sequencing began to use it to keep track of w h o had written w h i c h program,
equipment, most of it built specially for or by CERN. w h i c h program ran on w h i c h machine, w h o was part of w h i c h
At this time, a computer was still a sort of shrine to w h i c h project. Informal discussions at CERN w o u l d invariably be
scientists and engineers made pilgrimage. Most people at CERN accompanied by diagrams of circles and arrows scribbled on nap-
did not have computer terminals i n their offices; they had to kins and envelopes, because it was a natural way to show relation-
come to a central facility, such as the terminal room next to the ships between people and equipment. I wrote a four-page manual
9
 t a n g l e s , !
i n k b , arc w e b s


for Enquire that talked about circles and arrows, and how useful one page w o u l d have thousands of links on it that the page's
it was to use their equivalent in a computer program. owner might not want to bother to store. Furthermore, if an exter-
In Enquire, I could type in a page of information about a per- nal link went i n both directions, then changing both files w o u l d
son, a device, or a program. Each page was a "node" in the pro- involve storing the same information in two places, w h i c h is
gram, a little like an index card. The only way to create a new almost always asking for trouble: the files would inevitably get out
node was to make a link from an old node. The links f r o m and to of step.
a node w o u l d show up as a numbered list at the bottom of each Eventually, I compiled a database of people and a database of
page, much like the list of references at the end of an academic
software modules, but then m y consulting time was up. When I
paper. The only way of finding information was browsing from
left CERN, I didn't take the Enquire source code w i t h me. I had
the start page.
written it in the programming language Pascal, w h i c h was com-
I liked Enquire and made good use of it because it stored mon, but it ran on the proprietary Norsk Data SINTRAN-III oper-
information without using structures like matrices or trees. The ating system, w h i c h was pretty obscure. I gave the eight-inch
human m i n d uses these organizing structures all the time, but floppy disk to a systems manager, and explained that it was a pro-
can also break out of them and make intuitive leaps across the gram for keeping track of information. I said he was welcome to
boundaries â€"those coveted random associations. Once I discov- use it if he wanted. The program was later given to a student,
ered such connections, Enquire could at least store them. As I who said he liked the way it was w r i t t e n - w r i t t e n as a Pascal
expanded Enquire, I kept a vigilant focus on maintaining the con-
program should be w r i t t e n . The few people w h o saw it thought it
nections I was making. The program was such that I could enter
was a nice idea, but no one used it. Eventually, the disk was lost,
a new piece of knowledge only if I linked it to an existing one.
and w i t h it, the original Enquire.
For every link, I had to describe what the relationship was. For
W h e n I left CERN I rejoined a former colleague, John Poole.
example, if a page about Joe was linked to a page about a pro-
Two years earlier, Kevin and I had been working w i t h John, try-
gram, I had to state whether Joe made the program, used it, or
ing to upgrade the then-boring dot matrix printers w i t h the then-
whatever. Once told that Joe used a program, Enquire w o u l d also
revolutionary microprocessor so they could print fancy graphics.
know, w h e n displaying information about the program, that it
The three of us w o u l d sit i n the front room of John's house, his
was used by Joe. The links worked both ways.
golden Labrador nestled under one of the desks, and t r y to per-
Enquire ran on the group's software development computer. fect the design. We had succeeded in just a few months, but John
It did not r u n across a network, and certainly not the Internet,
hadn't had the money to go on paying us a salary, and w o u l d n ' t
w h i c h w o u l d not be used at CERN for years to come. Enquire
until he'd sold the product. That's w h e n we had started looking
had two types of links: an "internal" link f r o m one page (node)
for contract w o r k and ended up at CERN.
to another i n a file, and an "external" l i n k that could j u m p
After I had been at CERN for six months, John called. "Why
between files. The distinction was critical. A n internal l i n k
don't you come back?" he said. "I've sold the product, we've got a
w o u l d appear on both nodes. A n external link went i n only one
contract. N o w we need some software support for i t . " John had
direction. This was important because, if many people w h o were
incorporated as Image Computer Systems, and Kevin and I returned
making such a l i n k to one page could impose a return link, that
to help.
11
 w e a v i n g t h e w e b t a n g l e s , l i n k s , a n d w e b s


We rewrote all the motor controls to optimize the movement would create a node that represented the sequence. Whenever
of the print head so it was fast. It could also print Arabic, draw the same sequence occurred again, instead of repeating it, Tangle
three-dimensional pictures, and give the effect of preprinted sta- just put a reference to the original node. As more phrases were
tionery w h i l e using less expensive paper. We wrote our o w n stored as nodes, and more pointers pointed to them, a series of
markup language i n w h i c h documents were prepared, and the connections formed.
printer could also handle input codes of much more expensive The philosophy was: What matters is i n the connections. It
typesetting machines. We could change not only fonts but almost isn't the letters, it's the way they're strung together into words.
any aspect of the printer's behavior. It isn't the words, it's the way they're strung together into
The business went well, but the technology we were working phrases. It isn't the phrases, it's the way they're strung together
w i t h was limited to what we could put into printers. I felt I needed into a document. I imagined putting i n an encyclopedia this way,
a change f r o m living i n Britain, and I remembered that CERN had then asking Tangle a question. The question w o u l d be broken
a fellowship program. I n the spring of 1983 I decided to apply, down into nodes, w h i c h w o u l d then refer to wherever the same
arriving eventually i n September 1984. As a gift upon m y depar- nodes appeared i n the encyclopedia. The resulting tangle w o u l d
ture f r o m Image, John gave me a Compaq personal computer. It contain all the relevant answers.
was touted as one of the first "portable" computers, but it looked I tested Tangle by putting i n the phrase "How much wood
more like a sewing machine, more "luggable" than portable. W i t h would a woodchuck chuck?" The machine thought for a bit and
my new PC, and the freshness that comes w i t h change, I wrote i n encoded my phrase i n what was a very complex, tangled data struc-
my spare time another play program, called Tangle. I wanted to ture. But when I asked it to regurgitate what it had encoded, it
continue to explore the ideas about connections that were evolv- would follow through all the nodes and output again, "How much
ing i n m y head. wood would a woodchuck chuck?" I was feeling pretty confident,
so I tried it on "How much wood would a woodchuck chuck if a
I n an extreme view, the w o r l d can be seen as only connections, woodchuck could chuck wood?" It thought for a while, encoded it,
nothing else. We think of a dictionary as the repository of mean- and when I asked it to decode, it replied: "How much wood would
ing, but it defines words only i n terms of other words. I liked the a woodchuck chuck if a woodchuck chuck wood chuck chuck
idea that a piece of information is really defined only by what it's chuck wood wood chuck chuck chuck . . ." and it went on forever.
related to, and how it's related. There really is little else to mean- The mess it had made was so horrendously difficult to debug that I
ing. The structure is everything. There are billions of neurons i n never touched it again. That was the end of Tangleâ€"but not the
our brains, but what are neurons? Just cells. The brain has no end of m y desire to represent the connective aspect of information.
knowledge u n t i l connections are made between neurons. A l l that I had always stayed on the boundary of hardware and soft-
we know, all that we are, comes f r o m the way our neurons are ware, w h i c h was an important and exciting place to be, especially
connected. as software more and more took over hardware functions. W h e n
Computers store information as sequences of characters, so I applied for m y fellowship to CERN, I specified that I wanted a
meaning for them is certainly i n the connections among charac- job that w o u l d allow me to w o r k on both, and suggested three
ters. I n Tangle, if a certain sequence of characters recurred, it places there where I could do that. I ended up being hired to
12 13
 w e a v i n g t h e w e b t a n g l e s , l i n k s , a n d w e b s


w o r k w i t h "data acquisition and control," the group responsible I began to re-create Enquire on the Compaq. I wrote the pro-
for capturing and processing the results of experiments. Peggie gram so that it w o u l d r u n on both the luggable Compaq and the
Rimmer, w h o hired me, w o u l d also teach me, as it turned out, a VAX minicomputer made by DEC that I was using at CERN. I
lot about w r i t i n g standards, w h i c h was to come i n useful later didn't do such a good job the second time around, though: I just
on. I was i n a position to see more of CERN this time, to appreci- programmed i n the internal links, and never got around to w r i t -
ate more of its complexity. Although attached to a central com- ing the code for the external links. This meant that each web was
puting division, m y group worked w i t h the individual experiment limited to the notes that w o u l d fit i n one file: no link could con-
groups, each of w h i c h was a diverse mixture of scientists f r o m all
nect those closed worlds. The debilitating nature of this restric-
over the w o r l d .
tion was an important lesson.
By 1984, CERN had grown. A new accelerator, the Large Elec- It was clear to me that there was a need for something like
tron Positron accelerator, was being built. Its tunnel, twenty-seven Enquire at CERN. I n addition to keeping track of relationships
kilometers i n circumference, ran f r o m a hundred meters under between all the people, experiments, and machines, I wanted to
CERN to, at its farthest point, three hundred meters beneath the access different kinds of information, such as a researcher's tech-
foothills of the Jura mountains, dwarfing other accelerators. The nical papers, the manuals for different software modules, m i n -
computing diversity had increased too. A newer generation of utes of meetings, hastily scribbled notes, and so on. Furthermore,
computers, operating systems, and programming languages was I found myself answering the same questions asked frequently of
being used, as were a variety of networking protocols to link the me by different people. It w o u l d be so m u c h easier if everyone
many computers that sustained the big experiments. Machines could just read my database.
f r o m I B M , Digital Equipment Corp. (DEC), Control D a t a - w e What I was looking for fell under the general category of docu-
had them all, as w e l l as the new choice of PC or Mac i n personal mentation systemsâ€" software that allows documents to be stored
computers and different w o r d processors. and later retrieved. This was a dubious arena, however. I had seen
People brought their machines and customs w i t h them, and numerous developers arrive at CERN to tout systems that "helped"
everyone else just had to do their best to accommodate them. people organize information. They'd say, "To use this system all
Then teams went back home and, scattered as they were across you have to do is divide all your documents into four categories" or
time zones and languages, still had to collaborate. I n all this con- "You just have to save your data as a WordWcnderful document" or
nected diversity, CERN was a microcosm of the rest of the w o r l d , whatever. I saw one protagonist after the next shot down i n flames
though several years ahead i n time. by indignant researchers because the developers were forcing
I wrote a general "remote procedure call" (RPC) program to them to reorganize their w o r k to fit the system. I w o u l d have to
facilitate communication between all the computers and net- create a system w i t h common rules that w o u l d be acceptable to
works. W i t h RPC, a programmer could write a program on one everyone. This meant as close as possible to no rules at all.
sort of computer but let it call procedures on other computers, This notion seemed impossible until I realized that the diver-
even i f they ran on different operating systems or computer lan- sity of different computer systems and networks could be a rich
guages. The RPC tools w o u l d w o r k over whatever network or resource-something to be represented, not a problem to be eradi-
cable there happened to be available i n a given case. cated. The model I chose for m y minimalist system was hypertext.

15
 w e a v i n g t h e w e b t a n g l e s , l i n k s , a n d w e b s


M y vision was to somehow combine Enquire's external links B v late 1983 I was plotting to somehow get a hypertext system
w i t h hypertext and the interconnection schemes I had developed o i n a I talked to m y boss, Mike Sendall. He said it sounded like
for RPC. A n Enquire program capable of external hypertext links a reasonable idea, but that I should write up a proposal. A pro-
was the difference between imprisonment and freedom, dark and osal? I had no idea what went into a "proposal" at CERN. I
light. N e w webs could be made to bind different computers thought, however, that I ' d never get the go-ahead to develop a
together, and all new systems w o u l d be able to break out and ref- hypertext documentation system unless it was approved as a for-
erence others. Plus, anyone browsing could instantly add a new mal project. I thought hard about how to get the excitement of
node connected by a new link. this idea into a f o r m that w o u l d convince people at CERN.
The system had to have one other fundamental property: It Although Enquire provided a way to link documents and
had to be completely decentralized. That w o u l d be the only way databases, and hypertext provided a common format i n w h i c h to
a new person somewhere could start to use it w i t h o u t asking for display them, there was still the problem of getting different
access f r o m anyone else. A n d that w o u l d be the only way the sys- computers w i t h different operating systems to communicate w i t h
tem could scale, so that as more people used it, it w o u l d n ' t get each other. Ben Segal, one of my mentors i n the RPC project, had
bogged down. This was good Internet-style engineering, but most worked i n the States and had seen the Internet. He had since
systems still depended on some central node to w h i c h everything become a lone evangelist for using it at CERN. He went around
had to be c o n n e c t e d - a n d whose capacity eventually limited the pointing out how Unix and the Internet were binding universities
growth of the system as a whole. I wanted the act of adding a and labs together all over America, but he met a lot of resistance.
new link to be trivial; if it was, then a web of links could spread The Internet was nearly invisible i n Europe because people there
evenly across the globe. were pursuing a separate set of network protocols being designed
So long as I didn't introduce some central link database, and promoted by the International Standards Organization (ISO).
everything w o u l d scale nicely. There w o u l d be no special nodes, Whether because of the "not invented here" feeling, or for honest
no special links.. A n y node w o u l d be able to link to any other technical reasons, the Europeans were trying to design their o w n
node. This w o u l d give the system the flexibility that was needed, international network by committee.
and be the key to a universal system. The abstract document I was intrigued w i t h the Internet, though. The Internet is a
space it implied could contain every single item of information very general communications infrastructure that links computers
accessible over n e t w o r k s - a n d all the structure and linkages together. Before the Internet, computers were connected using
between them. dedicated cables f r o m one to another. A software program on one
Hypertext would be most powerful i f it could conceivably computer w o u l d communicate over the cable w i t h a software
point to absolutely anything. Every node, d o c u m e n t - w h a t e v e r it program on another computer, and send information such as a
was c a l l e d - w o u l d be fundamentally equivalent i n some way. hie or a program. This was originally done so that the very
Each w o u l d have an address by w h i c h it could be referenced. expensive early computers i n a lab or company could be used
They w o u l d all exist together i n the same s p a c e - t h e information f r o m different sites. Clearly, though, one computer could not be
space. linked to more than a few others, because it w o u l d need tens or
hundreds of cables running f r o m i t .
16 17
 w e a v i n g t h e w e b t a n g ' e s , l i n k s , a n d w e b s



The solution was to communicate indirectly over a network. I was interested i n the Internet because it could perhaps
The Internet is a network of networks. Its essence, though, is a •ide a bridge between different computer operating systems
set of standardized protocols-conventions by w h i c h computers
PT
°d networks. CERN was a technological melting pot. M a n y
0




send data to each other. The data are transmitted over various *" s i d s t s were used to Digital's VAX/VMS operating system
carriers, such as telephone lines, cable TV wires, and satellite P
J the DECnet communications protocols. Others preferred
channels. The data can be text, an e-mail message, a sound, an the - r o w i n g rival operating system, Unix, w h i c h used Internet
image, a software p r o g r a m - w h a t e v e r . When a computer is rotocols Every time a new experiment got started there w o u l d
ready to send its data, it uses special software to break the data be battles over whether to use VAX/VMS and DECnet, or U n i x
into packets that w i l l conform to two Internet protocols that gov- and TCP IP. I was beginning to favor TCP/IP myself, because
ern how the packets w i l l be shipped: IP [Internet Protocol) and TCP was starting to become available for the VMS, too. It didn't
TCP (Transmission Control Protocol). The software labels each initially come f r o m Digital, but f r o m Wollongong University i n
packet w i t h a unique number. It sends the packets out over the Australia. _
phone or cable wire, and the receiving computer uses its o w n
Using TCP/IP would mean that the U n i x world, w h i c h already
Internet software to put them back together according to the
used TCP/IP, w o u l d be satisfied, and those i n the VAX w o r l d could
labels.
get into the' Unix world, too. Finally, there was a way for both
The Internet was up and running by the 1970s, but transfer- contenders to communicate w i t h each other, by picking up a piece
ring information was too much of a hassle for a noncomputcr of TCP/IP software from Wollongong. I became so convinced
expert. One would r u n one program to connect to another com- about TCP/IP's significance that I added code to the RPC system
puter, and then i n conversation (in a different language) w i t h the so that it could communicate using TCP/IP, and created an
other computer, r u n a different program to access the informa- addressing system for it that identified each remote service i n the
tion. Even w h e n data had been transferred back to one's own RPC system. That's w h e n the Internet came into m y life.
computer, decoding it might be impossible. For the proposal, I also had to think out what was needed to
Then electronic mail was invented. E-mail allowed messages scale up Enquire into a global system. I w o u l d have to sell this
to be sent f r o m one person to another, but it did not f o r m a space project as a documentation s y s t e m - a perceived need at C E R N -
in w h i c h information could permanently exist and be referred to. and not as a hypertext system, w h i c h just sounded too precious.
Messages were transient. (When the World Wide Web arrived, But if this system was going to go up as a way of accessing infor-
riding on top of the Internet, it w o u l d give information a place to
mation across a network, it w o u l d be i n competition w i t h other
persist.)
documentation systems at CERN. Having seen prior systems shot
CERN's lateness i n adopting the Internet was surprising, down, I knew the key w o u l d be to emphasize that it w o u l d let
because the laboratory had been very much on the leading edge each person retain his o w n organizational style and software on
of networking and telecommunications. It had developed CERN¬ his computer.
net, its own home-brewed network, for lack of commercial net- The system needed a simple way for people to represent links
works. It had its own e-mail systems. A n d it was at the forefront in their documents, and Lo navigate across links. There was a
of gatewaying between different proprietary mail and file systems. model i n online "help'' programs: If there was an instruction or
iB 19
 t a n 5 1
e s , ' i n k s , a n d w e b s
w e a v i n g t h e w e b


tool on the screen that a user didn't understand, he just clicked In March 1989 I took the leap to write a proposal. I wanted to

on it and more information would appear. This approach was explain that generality was the essence of a web of information.

called hot buttons, a derivative of Ted Nelson's hypertext that had On the other hand, I felt I had to make the system seem to be

subsequently been used by Apple Computer's "HyperCard" and something that could happen only at CERN. I was excited about
later in some way by many point-and-click help systems. I decided escaping f r o m the straitjacket of hierarchical documentation sys-
that on m y system, if someone wanted to put a hypertext link tems, but I didn't want the people responsible for any hierarchi-
into a piece of text, the words noting the link w o u l d be high- cal system to throw rocks at me. I had to show how this system
lighted in some way on the screen. If a viewer clicked on a high- could integrate very disparate things, so I provided an example of
lighted word, the system would take h i m to that link. an Internet newsgroup message, and a page f r o m m y old Enquire
program.
The pieces were starting to fall into place. TCP/IP would be the I was brash enough to look forward to having a web of data
network protocol of choice. For "marketing" purposes, I would that could be processed by machine. I said:
propose the system as one that would w o r k over DECnet, w i t h
the added benefit that someone could communicate over the An intriguing possibility, given a large hypertext database
Internet, too. That left one hole: For people to communicate and with typed links, is that it allows some degree of automatic
share documents, they had to have a simple but common analysis. [ . . . ] Imagine making a large three-dimensional
addressing scheme so they'd know how to address their files and model, with people represented by little spheres, and strings
others w o u l d know how to request files. I adapted the simple between people who have something in common at work.
RPC addressing scheme. Now imagine picking up the structure and shaking it,

In presenting my argument to an experiment group, I would until you make some sense of the tangle: Perhaps you see

note that they typically have different kinds of documented infor- tightly knit groups in some places, and in some places weak

mationâ€"a "help" program, a telephone book, a conference infor- areas of communication spanned by only a few people. Per-

mation system, a remote library system â€"and they w o u l d be haps a linked information system w i l l allow us to see the

looking for ways to create a consistent master system. They real structure of the organization in which we work.

w o u l d have three choices: ¡1) design yet another documentation
scheme that is supposedly better than all the ones that have been Little d i d I k n o w that Ph.D. theses w o u l d later be done on such
attempted before it; (2) use one of the existing schemes and make topics.
do w i t h its limitations; or (3) realize that all these remote systems For all the decisions about w h i c h technical points to include
have something in common. I w o u l d tell them, "We can create a in the proposal or exclude, and w h i c h social advantages of the
common base for communication while allowing each system to system to emphasize, I was rather light on the project manage-
maintain its individual*-/. That's what this proposal is about, and ment details:
global hypertext is what w i l l allow you to do it. A l i you have to
do is make up an address for ..-arh document or screen in your I imagine that two people for six to twelve months would be
system and the rest is easy." sufficient for this phase of the project. A second phase would
?i
 w e a v i n g t h e w e b t a n g l e s , l i n k s , a n d w e b s


almost certainly involve some programming in order to set t ly been started by Steve Jobs, who had founded Apple
up a real system at CERN on many machines. An important r e c e n
u t e r a n d brought the first intuitive point-and-click, folders
part of this, discussed below, is the integration of a hypertext C0
^ f a c e to personal computers. Ben Segal, our Unix and Inter-
system with existing data, so as to provide a universal sys- inter
v a n g e l i s t , had mentioned that the N e X T machine had a lot
tem, and to achieve critical usefulness at an early stage. "f Uriguing features that might help us. I asked Mike to let me
° 1
one (bringing Ben with me for weight), and he agreed. He
By the end of March 1989 I had given the proposal to Mike also said, "Once you get the machine, why not try programming
Sendall; to his boss, David Williams; and to a few others. I gave it a
o u r hypertext thing on it?" I thought I saw a twinkle in his eye.
to people at a central committee that oversaw the coordination of Y
° By buying a NeXT, we could justify my working on my long-
computers at C E R N . But there was no forum from which I could delayed hypertext project as an experiment in using the NeXT
command a response. Nothing happened. operating system and development environment. I immediately
While I waited for some kind of feedback, I tested the idea in began to think of a name for my nascent project. I was looking
conversation, and reactions varied. C E R N people moved through " for words that would suggest its new kind of structure. Mesh, or
a number of overlapping loyalties, perhaps one to C E R N , one to Information Mesh, was one idea (used in the diagram in the pro-
an experiment, to an idea, to a way of doing things, to their origi- posal), but it sounded a little too much like mess. I thought of
nal institute . . . not to mention the set of Macintosh users or Mine of Information, or MOI, but moi in French means "me," and
IBM/PC users. Another reason for the lackluster response was that was too egocentric. An alternative was The Information Mine,
that C E R N was a physics lab. There were committees to decide but that acronym, T I M , was even more egocentric! Besides, the
on appropriate experiments, because that was the stock-in-trade, idea of a mine wasn't quite right, because it didn't encompass the
but information technology was very much a means to an end, idea of something global, or of hypertext, and it represented only
with less structure to address it. The situation was worse for very getting information outâ€"not putting it in.
general ideas such as global hypertext. Even the R P C project, also I was also looking for a characteristic acronym. I decided that
an exercise in generality, had little formal support from within I would start every program involved in this system with "HT,"
C E R N , but it had enough support among different groups that I for hypertext. Then another name came up as a simple way of
could keep it going. representing global hypertext. This name was used in mathemat-
In the meantime, I got more involved with the Internet, and ics as one way to denote a collection of nodes and links in which
read up on hypertext. That's when I became more convinced any node can be linked to any other. The name reflected the dis-
than ever that I was on the right track. By early 1990 I still had tributed nature of the people and computers that the system
received no reactions to the proposal. I decided to try to spark could link. It offered the promise of a potentially global system.
some interest by sending it around again. I reformatted it and put Friends at C E R N gave me a hard time, saying it would never
a new date on it: May 1990. I gave it to David Williams again, take off-especially since it yielded an acronym that was nine
and again it got shelved. syllables long when spoken. Nonetheless, I decided to forge
During this time I was talking to Mike Sendall about buying a ahead. I would call my system the "World Wide Web."
new kind of personal computer called the NeXT. N e X T Inc. had
23
 r



C H A P T E R




i n f o . c e r n . c h




W h i l e it seemed to be uphill work convincing anyone at C E R N
that global hypertext was exciting, one person was an immediate
convert: Robert Cailliau.
Though now the Electronics and Computing for Physics
division, by coincidence Robert had in 1980 been in the same
Proton Synchotron division as I, and had in fact written the
text-formatting program I had used to print the Enquire manual.
A Flemish-speaking Belgian, Robert had had the lifelong frustra-
tion of people insisting on addressing him in French. After taking
an engineering degree at the University of Ghent he picked up a
master's at the University of Michigan, an experience that left
him with an accent in English that is impossible to identify.
Indeed, it became a parlor game for newcomers at C E R N to try
to guess exactly where he was from.
A dapper dresser who methodically schedules haircuts
according to the solstice and equinox, Robert is fastidious in all


III
w e a v i n g t h e w e b i n F o . c e r n . c h J




things. He is the kind of engineer who can be driven mad by the that would open and display documents, and preferably let people
incompatibility of power plugs. No wonder, then, that he would edit them, too. All that was missing was the Internet. They've
be attracted to a solution to computer incompatibility, especially already done the difficult bit! I thought, so I tried to persuade them
coming with a simple user interface. In the marriage of hypertext to add an Internet connection. They were friendly enough, but
and the Internet, Robert was best man.
they, too, were unconvinced.
Robert's real gift was enthusiasm, translated into a genius for I got the same response from others at the conference. It
spreading the gospel. While I sat down to begin to write the seemed that explaining the vision of the Web to people was
Web's code, Robert, whose office was a several-minute walk exceedingly difficult without a Web browser in hand. People had
away, put his energy into making the W W W project happen at to be able to grasp the Web in full, which meant imagining a
C E R N . He rewrote a new proposal in terms he felt would have whole world populated with Web sites and browsers. They had to
more effect. A C E R N veteran since 1973, he lobbied among his sense the abstract information space that the Web could bring
wide network of friends throughout the organization. He looked
into being. It was a lot to ask.
for student helpers, money machines, and office space.
The hypertext community may also have been slightly demor-
By the time Mike Sendall approved my purchase of the NeXT alized. Their small conference was not getting any bigger, and no
machine, I had already gone to the hypertext industry looking for one was sure where the field was headed. The lack of commer-
products onto which we could piggyback the Web. At C E R N there cial successes had perhaps left a certain cynicism about bright,
was a "Buy, don't build" credo about acquiring new technology. new ideas that could change the world.
There were several commercial hypertext editors, and I thought Another possibility I saw was called Dynatext, and was from
we could just add some Internet code so the hypertext documents Electronic Book Technology, a company in Rhode Island started
could be sent over the Internet. I thought the companies engaged by Andy Van Dam, the Brown University researcher who had
in the then fringe field of hypertext products would immediately coined the term electronic book. I thought the company's software
grasp the possibilities of the Web. Unfortunately, their reaction could be turned into a Web browser/editor rather easily. However,
was quite the opposite. "Nope," they said. "Too complicated." like many hypertext products at the time, it was built around the
Undaunted, in September 1990 Robert and I went to the idea that a book had to be "compiled" (like a computer program)
European Conference on Hypertext Technology (ECHT) at Ver- to convert it from the form in which it was written to a form in
sailles to pitch the idea. The conference exhibition was small, but which it could be displayed efficiently. Accustomed to this cum-
there were a number of products on display, such as a multi- bersome multistep process, the E B T people could not take me
media training manual for repairing a car. seriously when I suggested that the original coded language could
I approached Ian Ritchie and the folks from O w l Ltd., which be sent across the Web and displayed instantly on the screen.
had a product called Guide. In Peter Brown's original Guide They also insisted on a central link database to ensure that
work at the University of Southampton, when a user clicked on a there were no broken links. Their vision was limited to sending
hypertext link, the new document would be inserted right there text that was fixed and consistentâ€"in this case, whole books. I
in place. The version now commercialized by Owl looked astonish- was looking at a living world of hypertext, in which all the pages
ingly like what I had envisioned for a Web browserâ€"the program would be constantly changing. It was a huge philosophical gap.
26 27
 w e a v i n g t h e w e b ¡ n f o . c e r n . c h



Letting go of that need for consistency was a crucial design step Protocol (HTTP), the language computers would use to com-
that would allow the Web to scale. But it simply wasn't the way municate over the Internet, and the Universal Resource Identifier
things were done. ,rjRl) the scheme for document addresses.
By mid-November I had a client p r o g r a m - a point-and-click
Despite the "Buy, don't build" credo, I came to the conclusion bro w s e r / e d i t o r - w h i c h I just called WorldWideWeb. By December
that I was going to have to create the Web on my own. In Octo- i t was working with the Hypertext Markup Language (HTML) I
ber 1990 I began writing code for the Web on my new computer h a d written, which describes how to format pages contaming
The N e X T interface was beautiful, smooth, and consistent. It had hypertext links. The browser would decode URIs, and let me read,
great flexibility, and other features that would not be seen on PCs write, or edit Web pages in H T M L . It could browse the Web
till later, such as voice e-mail, and a built-in synthesizer. It also using HTTP, though it could save documents only into the local
had software to create a hypertext program. Its failure to take computer system, not over the Internet.
over the industry, despite all these advantages, became for me a I also wrote the first Web server-the software that holds Web
cautionary tale. N e X T required users to accept all these innova-
pages on a portion of a computer and allows others to access
tions at onceâ€"too much.
them Like the first client, the server actually ran on my desktop
My first objective was to write the Web client-the program NeXT machine. Though the server was formally known as
that would allow the creation, browsing, and editing of hypertext ntt)c01.cern.ch (NeXT, Online Controls, 1), I registered an alias
pages. It would look basically like a word processor, and the tools for i t - " i n f o . c e r n . c h . " - w i t h the C E R N computer system folks.
on the NeXT's system, called NeXTStep, were ideal for the task. I That way, the server would not be tied by its address to my
could create an application, menus, and windows easily, just drag- NeXT machine; if I ever moved its contents to another machine,
ging and dropping them into place with a mouse. The meat of it all the hypertext links pointing to it could find it. I started the
was creating the actual hypertext window. Here I had some cod- first global hypertext Web page, on the info.cern.ch server, with
ing to do, but I had a starting place, and soon had a fully func- my own notes, specifications of HTTP, U R I , and H T M L , and all
tional word processor complete with multiple fonts, paragraph the project-related information.
and character formatting, even a spellchecker! No delay of gratifi-
At this point Robert bought his own NeXT machine and we
cation here. Already I could see what the system would look like.
reveled in being able to put our ideas into practice: communica-
I still had to find a way to turn text into hypertext, though
tion through shared hypertext.
This required being able to distinguish text that was a link from
At long last I could demonstrate what the Web would look
text that wasn't. I delved into the files that defined the internal
like. But it worked on only one platform, and an uncommon one
workings of the text editor, and happily found a spare thirty-two-
at t h a t - t h e NeXT. The H T T P server was also fairly crude. There
bit piece of memory, which the developers of N e X T had gra-
was a long way to go, and we needed help. Ben Segal, who had a
ciously left open for future use by tinkerers like me. I was able to
knack for adjusting staffing levels behind the scenes, spotted a
use the spare space as a pointer from each span of text to the
young intern named Nicola Pellow. A math student from England,
address for any hypertext link. With this, hypertext was easy. I
Nicola was working for a colleague in a neighboring building but
was then able to rapidly write the code for the Hypertext Trans-
didn't have enough to do.
28 29
w e a v i n g t h e w e b i n f o . c e r n . c h


A big incentive for putting a document on the Web was that child, due Christmas Eve. As fate would have it, she waited a few
anyone else in the world could find it. But who would bother to extra days. We drove to the hospital during a New Year's Eve
install a client if there wasn't exciting information already on the storm and our daughter was born the next day. As amazing as it
Web? Getting out of this chicken-and-egg situation was the task would be to see the Web develop, it would never compare to see-
before us. We wanted to be able to say that if something was on ing the development of our child.
the Web, then anyone could have access to itâ€"not just anyone
with a NeXT! As the new year unfolded, Robert and I encouraged people in the
When I gave talks, I showed a diagram with machines of all Computing and Networking division to try the system. They didn't
types connected to the Internet, from mainframes with simple seem to see how it would be useful. This created a great tension
character-oriented terminals through PCs, Macs, arid more. To among us about how to deploy our limited resources. Should we
make this possible, I urged Nicola to give the Web the best be evangelizing the Web? Should we develop it further on the
browser she could, but to assume as little as possible, so this NeXT? Should we reprogram it for the Mac or the P C or Unix,
interface could work on any kind of computer. The least common because even though the N e X T was an efficient machine, few
denominator we could assume among all different types of com- other people had them? After all, what good was a "worldwide"
puters was that they all had some sort of keyboard input device, web if there were only a few users? Should we tailor the Web to
and they all could produce ASCII (plain text) characters. The the high-energy physics community, so they'd have a tool that
browser would have to be so basic that it could even work on a was theirs and would support it, since C E R N was paying our
paper Teletype. We therefore called it a line-mode browser, salaries? Or should we generalize the Web and really address the
because Teletype machines and the earliest computer terminals global community, at the risk of being personally disenfranchised
operated by displaying text one line at a time. by C E R N ? .
Meanwhile, I took one quick step that would demonstrate the Trading in the NeXT for some ordinary computer would have
concept of the Web as a universal, all-encompassing space. I pro- been like trading in a favorite sports car for some truck. More
grammed the browser so it could follow links not only to files on important, the Web was already written for it. If we switched to
H T T P servers, but also to Internet news articles and newsgroups. developing the Web for the much more widely used P C , accep-
These were not transmitted in the Web's H T T P protocol, but in an tance might be quicker, but the point was to get people to try
Internet protocol called F T P (file transfer protocol). With this what we already had. If we stopped progress and went back to
move, Internet newsgroups and articles were suddenly available redoing things for the PC, we might never get it done. I decided
as hypertext pages. In one fell swoop, a huge amount of the infor- to stick with the NeXT.
mation that was already on the Internet was available on the Web. As for the application, my gut told me I had to pursue my
The WorldWideWeb browser/editor was working on my machine larger vision of creating a global system. My head reminded me,
and Robert's, communicating over the Internet with the info.cern.ch however, that to attract resources I also needed a good, visible
server by Christmas Day 1990. reason to be doing this at C E R N . I was not employed by C E R N to
As significant an event as this was, I wasn't that keyed up create the Web. At any moment some higher-up could have ques-
about it, only because my wife and I were expecting our first tioned how I was spending my time, and while it was unusual to
30 31
 ¡ n f o . c e - r n . c h
w e a v i n g t h e w e b


stop people at C E R N from following their own ideas, my infor- it left people thinking of the Web as a medium in which a few
mal project could have been ended. However, it was too soon to published and most browsed. My vision was a system in which
try to sell the Web as the ultimate documentation system that sharing what you knew or thought should be as easy as learning
would allow all of C E R N ' s documents, within and between proj- what someone else knew.
ects, to be linked and accessible, especially given the history of so Mundane as it was, this first presentation of the Web was, in
many failed documentation systems. Small but quantifiable steps a curious way, a killer application. Many people had workstations,
seemed in order. Our first target, humble beginning that it was, with one window permanently logged on to the mainframe just
would be the C E R N telephone book. to be able look up phone numbers. We showed our new system
The phone book existed as a database on C E R N ' s aging main- around C E R N and people accepted it, though-most of them didn't
frame. Bernd Pollermann, who maintained it and all sorts of understand why a simple ad hoc program for getting phone num-
other central information, was charged with somehow providing bers wouldn't have done just as well.
all this material to each and every user on his or her favorite sys- Of course, we didn't want our brainchild with all its tremen-
tem. I managed to persuade Bernd that the Web was just what he dous potential to be locked in at this rather pedestrian level. To
needed to make life a great deal simpler. If he created a server, I broaden the Web's horizons, I set about giving talks and conduct-
told him, we would get the browsers onto everyone's desktop. He ing demonstrations. So that people could see something "out
went for it. there on the Web" other than the phone book, and to practice
I got my simple server to run on the mainframe, then chopped what we preached, Robert and I continued to document the proj-
it in two, so that the essential HTTP-related Internet functions ect in hypertext on info.cern.ch.
were done by my code (written in C language) and Bernd was left What we had accomplished so far was based on a few key
to write the rest of the server in his favorite language, "REXX." To principles learned through hard experience. The idea of univer-
make all the documents available, he just had to learn to write sality was key: The basic revelation was that one information
H T M L , which took him only a few afternoons. Soon the entire space could include them all, giving huge power and consistency.
world of his search engines, databases, and catalogues was avail- Many of the technical decisions arose from that. The need to
able as hypertext. encode the name or address of every information object in one
That brought us back to the search for a browser. We started URI string was apparent. The need to make all documents in

porting Nicola's line-mode client onto all sorts of machines, from some way "equal" was also essential. The system should not con-

mainframes through Unix workstations to plain DOS for the P C . strain the user; a person should be able to link with equal ease to

These were not great showcases for what the Web should look any document wherever it happened to be stored.

like, but we established that no matter what machine someone This was a greater revelation than it seemed, because hyper-
was on, he would have access to the Web. This was a big step, text systems had been limited works. They existed as databases
but it was achieved at some sacrifice in that we decided not to on a floppy disk or a CD-ROM, with internal links between their

take the time to develop the line-mode browser as an editor. Sim- hies. For the Web, the external link is what would allow it to

ply being able to read documents was good enough to bootstrap actually become "worldwide." The important design element

the process. It justified Bernd's time in getting his servers up. But would be to ensure that when two groups had started to use the

32 33
 w e a v i n g t h e w e b



Web completely independently at different institutions, a person C H A P T E R 4
m one g r o u p c o^ u l d ^
c r e a t e ^a t Q ^ g




with only a small incremental effort, and without having to
merge the two document databases or even have access to the
other system. If everyone on the Web could do this, then a single
hypertext link could lead to an enormous, unbounded world
P r o t o c o l s


Simple Rules For Global Systems




Incompatibility between computers had always been a huge
pain in everyone's side, at C E R N and anywhere else where they
were used. C E R N had all these big computers from different
manufacturers, and various personal computers, too. The real
world of high-energy physics was one of incompatible networks,
disk formats, data formats, and character-encoding schemes,
which made any attempt to transfer information between com-
puters generally impossible. The computers simply could not
communicate with each other. The Web's existence would mark
the end of an era of frustration.
As if that weren't advantage enough, the Web would also pro-
vide a powerful management tool. If people's ideas, interactions,
and work patterns could be tracked by using the Web, then'com-
puter analysis could help us see patterns in our work, and facili-
tate our working together through the typical problems that beset
any large organization.
One of the beautiful things about physics is its ongoing quest
to find simple rules that describe the behavior of very small,
simple objects. Once found, these rules can often be scaled up to
34
35
 w e a v i n g t h e w e b p r o t o c o l s


describe the behavior of monumental systems in the real world documents or graphics, they can share directly. If not, they can
For example, by understanding how two molecules of a g a s
b o t h translate to H T M L .
interact when they collide, scientists using suitable mathematics The fundamental principle behind the Web was that once
can deduce how billions of billions of gas moleculesâ€"say, the someone somewhere made available a document, database,
earth's atmosphere-will change. This allows them to analyze
graphic, sound, video, or screen at some stage in an interactive
global weather patterns, and thus predict the weather. If the
dialogue, it should be accessible (subject to authorization, of
rules governing hypertext links between servers and browsers
course) by anyone, with any type of computer, in any country.
stayed simple, then our web of a few documents could grow to a
And it should be possible to make a referenceâ€"a linkâ€"to that
global web.
thing, so that others could find it. This was: a philosophical
The art was to define the few basic, common rules of "proto- change from the approach of previous computer systems. People
col" that would allow one computer to talk to another, in such a
were used to going to find information, but they rarely made ref-
way that when all computers everywhere did it, the system
erences to other computers, and when they did they typically had
would thrive, not break down. For the Web, those elements were,
to quote a long and complex series of instructions to get it. Fur-
in decreasing order of importance, universal resource identifiers
thermore, for global hypertext, people had to move from thinking
(URIs), the Hypertext Transfer Protocol (HTTP), and the Hyper-
about instructions to thinking in terms of a simple identifier
text Markup Language (HTML).
stringâ€"a URIâ€"that contained all the essential details in a com-
What was often difficult for people to understand about the
pact way.
design was that there was nothing else beyond URIs, HTTP, and
Getting people to put data on the Web often was a question
H T M L . There was no central computer "controlling" the Web, no
of getting them to change perspective, from thinking of the user's
single network on which these protocols worked, not even an
access to it not as interaction with, say, an online library system,
organization anywhere that "ran" the Web. The Web was not a
but as navigation though a set of virtual pages in some abstract
physical "thing" that existed in a certain "place." It was a "space"
space. In this concept, users could bookmark any place and
in which information could exist.
return to it, and could make links into any place from another
I told people that the Web was like a market economy. In a document. This would give a feeling of persistence, of an ongoing
market economy, anybody can trade with anybody, and they existence, to each page. It would also allow people to use the
don't have to go to a market square to do it. What they do need, mental machinery they naturally have for remembering places
however, are a few practices everyone has to agree to, such as the and routes. By being able to reference anything with equal ease,
currency used for trade, and the rules of fair trading. The equiva- the Web could also represent associations between things that
lent of rules for fair trading, on the Web, are the rules about what might seem unrelated but for some reason did actually share a
a U R I means as an address, and the language the computers relationship. This is something the brain can do easily, sponta-
useâ€"HTTPâ€"whose rules define things like which one speaks neously. If a visitor came to my office at C E R N , and I had a fresh
first, and how they.speak in turn. When two computers agree cutting of lilac in the corner exuding its wonderful, pungent
they can talk, they then have to find a common way to represent scent, his brain would register a strong association between the
their data so they can share it. If they use the same software for office and lilac. He might walk by a lilac bush a day later in a
36 .
37
 w e a v i n g t h e w e b p r o t o c o l s


park and suddenly be reminded of my office. A single click: l i l a c Of course if I had insisted everyone use HTTP, this would
. . . office. also have been against the principle of minimal constraint. If the
The research community had used links between paper docu- y/gb were to be universal, it should be as unconstraining as pos-
ments for ages: Tables of contents, indexes, bibliographies, and sible. Unlike the N e X T computer, the Web would come as a set
reference sections are hypertext links. On the Web, however, of ideas that could be adopted individually in combination with
research ideas in hypertext links can be followed up in seconds' existing or future parts. Though H T T P was going to be faster,
rather than weeks of making phone calls and waiting for deliver- w h o was I to say that people should give up the huge archives of
ies in the mail. And suddenly, scientists could escape from the data accessible from F T P servers?
sequential organization of each paper and bibliography, to pick The key to resolving this was the design of: the U R I . It is the
and choose a path of references that served their own interest. most fundamental innovation of the Web, because it is the one
But the Web was to be much more than a tool for scientists. specification that every Web program, client or server, anywhere
For an international hypertext system to be worthwhile, of course, uses when any link is followed. Once a document had a U R I , it
many people would have to post information. The physicist would could be posted on a server and found by a browser.
not find much on quarks, nor the art student on Van Gogh, if Hidden behind a highlighted word that denotes a hypertext
many people and organizations did not make their information link is the destination document's U R I , which tells the browser
available in the first place. Not only that, but much information- where to go to find the document. A U R I address has distinct
from phone numbers to current ideas and today's m e n u - i s con- parts, a bit like the five-digit zip code used by the U . S . postal
stantly changing, and is only as good as it is up-to-date. That system. The. first three numbers in a zip code designate a cer-
meant that anyone (authorized) should be able to publish and cor- tain geographic regionâ€"a town, or part of a city or county. The
rect information, and anyone (authorized) should be able to read next two numbers define a very specific part of that r e g i o n -
it. There could be no central control. To publish information, it say, a few square blocks in a city. This gets the mail to a local
would be put on any server, a computer that shared its resources post office. Carriers from there use the street name or box num-
with other computers, and the person operating it defined who ber to finish the routing.
could contribute, modify, and access material on it. Information Slashes are used in a U R I address to delineate its parts. The
was read, written, or edited by a client, a computer program, such first few letters in the U R I tells the browser which protocol to
as a browser/editor, that asked for access to a server. use to look up the document, whether H T T P or F T P or one of a
Several protocols already existed for transferring data over small set of others. In the address http://www.foobar.com/docl,
the Internet, notably NNTP for Network News and F T P for files. the www.foobar.com identifies the actual computer server where
But these did not do the negotiating I needed, among other these documents exist. The docl is a specific document on the
things. I therefore defined HTTP, a protocol simple enough to be www.foobar.com server (there might be hundreds, each with a
able to get a Web page fast enough for hypertext browsing. The different name after the single slash). The letters before the double
target was a fetch of about one-tenth of a second, so there was no slash signify the communications protocol this server uses.
time for a conversation. It had to be "Get this document," and The big difference between the U R I and postal schemes,
"Here it is!" however, is that while there is some big table somewhere of all
38 39
 p r o t o c o l s
w e a v i n g t h e w e b


zip codes, the last part of the U R I means whatever the given 0 f meetings, and mail m e s s a g e s - i n short, 95 percent of daily life

server wants it to mean. It doesn't have to be a file name. It can ° r most people. Hence H T M L , the Hypertext Markup Language.
0




be a table name or an account name or the coordinates of a map I expected H T M L to be the basic warp and weft of the Web,
or whatever. The client never tries to figure out what it means: It b u t documents of all types-video, computer-aided design, sound,
just asks for it. This important fact enabled a huge diversity of animation, and executable programs-to be the colored threads
types of information systems to exist on the Web. And it allowed that would contain much of the content. It would turn out that
the Web to immediately pick up all the NNTP and F T P content HTML would become amazingly popular for the content as well.
from the Internet. H T M L is a simple way to represent hypertext. Once the U R I
of a document tells a browser to talk H T T P to the server, then
At the same time that I was developing the Web, several other
client and server have to agree on the format of the data they will
Internet-based information systems were surfacing. Brewster
share, so that it can be broken into packets both will understand.
Kahle at Thinking Machines had architected their latest powerful
If they both knew WordPerfect files, for example, they could
parallel processor. Now he saw a market for the big machines as
swap WordPerfect documents directly. If not, they could both try
search engines and designed the Wide Area Information Servers
to translate to H T M L as a default and send documents that way.
(WAIS) protocol to access them to form a system like the Web but
There were some basic design rules that guided H T M L , and some
without linksâ€"only search engines.
pragmatic, even political, choices. A philosophical rule was that
Clifford Newman at the Information Sciences Institute pro-
HTML should convey the structure of a hypertext document, but
posed his Prospero distributed file system as an Internet-based
not details of its presentation. This was the only way to get it to
information system, and Mark McCahill and colleagues at the
display reasonably on any of a very wide variety of different
University of Minnesota were developing a campus-wide infor-
screens and sizes of paper. Since I knew it would be difficult to
mation system called gopher, named for the university's mascot.
encourage the whole world to use a new global information sys-
To emphasize that all information systems could be incorporated
tem, I wanted to bring on board every group I could. There was a
into the Web, I defined two new U R I prefixes that could appear
family of markup languages, the standard generalized markup
before the double slashâ€""gopher:" and "wais:"â€"that would give
language (SGML), already preferred by some of the world's docu-
access to those spaces. Both systems took off much more quickly
mentation community and at the time considered the only poten-
than the Web and I was quite concerned at the time that they
tial document standard among the hypertext community. I
would suffocate it.
developed H T M L to look like a member of that family.
H T T P had a feature called format negotiation that allowed a
Designing H T M L to be based on SGML highlighted one of the
client to say what sorts of data format it could handle, and allow
themes of the development of the Web: the constant interplay
the server to return a document in any one of them. I expected
between the diplomatically astute decision and the technically
all kinds of data formats to exist on the Web. I also felt there had
clean thing to do. SGML used a simple system for denoting
to be one common, basic lingua franca that any computer would
instructions, or "tags," which was to put a word between angle
be required to understand. This was to be a simple hypertext lan-
brackets (such as < hi > to denote the main heading of a page), yet
guage that would be able to provide basic hypertext navigation,
it also had many obscure and strange features that were not well
menus, and simple documentation such as help files, the minutes
4i
 w e a v i n g t h e w e b p r o t o c o l s



understood. Nonetheless, at the time, the Web needed support an drafted a work plan for the Electronics and Computing for
understanding from every community that could become involved* W C
- c s division, where Robert was, to try to get funding from
and in many ways the SGML community provided valuable input" F h y S 1
but no one responded. Accordingly, while developing the
S G M L was a diplomatic choice at C E R N as well. SGML was" ^rm'ology and trying to promote it to our colleagues, we still
being used on C E R N ' s I B M machines with a particular set of tags Td to maintain a somewhat low profile.
that were enclosed in angle brackets, so H T M L used the sam e
The other problem we faced was simply the climate at
tags wherever possible. I did clean up the language a certain; CERN. There was a constant background of people promoting
amount, but it was still recognizable. I chose this direction so that ideas for new software systems. There was competition among
when a C E R N employee saw the angle brackets of H T M L , he or 1
stems created within the experiment groups themselves-soft-
she would feel, Yes, lean do that. In fact, H T M L was even easier
ware for running a physics experiment, but also for everything
to use than C E R N ' s version of SGML. The people promoting the
from handling electronic mail and organizing documents to run-
S G M L system at C E R N could possibly be powerful figures in the
ning the Coke machine. There was competition over which net-
choice of C E R N ' s future directions and I wanted them to feel
work to use, among them DECnet, the Internet, and whatever
happy about the Web.
home-brewed thing could be justified. With so many creative
I never intended H T M L source code (the stuff with the angle engineers and physicists in one place, innovations were constant.
brackets) to be seen by users. A browser/editor would let a user At the same time, C E R N obviously couldn't tolerate everybody
simply view or edit the language of a page of hypertext, as if he creating unique software for every function.
were using a word processor. The idea of asking people to write Robert and I had to distinguish our idea as novel, and one
the angle brackets by hand was to me, and I assumed to many, as that would allow C E R N to leap forward. Rather than parade in
unacceptable as asking one to prepare a Microsoft Word docu- with our new system for cosmic sharing of information, we
ment by writing out its binary coded format. But the human decided to try to persuade people that we were offering them a
readability of H T M L was an unexpected boon. To my surprise, way to extend their existing documentation system. This was a
people quickly became familiar with the tags and started writing concrete and potentially promising notion. We could later get
their own H T M L documents directly. them to sign on to the dream of global hypertext. Our argument
was that everyone could continue to store data in any form they
As the technical pieces slowly fell into place, Robert and I were like, and manage it any way they like. The Web would simply
still faced with a number of political issues that gave us more help people send and access information between each other,
than a twinge of anxiety. First of all, the Web was still not a for- regardless of the operating system or formats their computers
mal project. At any moment some manager of the Computing and use. The only thing they'd have to do was follow the same simple
Networking division could have asked me to stop the work, as it
URI addressing scheme. They didn't "have to" use H T T P or
wasn't part of any project, and it could have been considered
HTML, but those tools were there if they ran into an incompati-
inappropriate for C E R N .
bility problem.
For eight months Robert, Nicola, and I refined the basic As we made these points, we also noted that using H T M L
pieces of the Web and tried to promote what we were creating. was easy, since it was so much like SGML. I may have promoted
43
weavirç the web

this angle too much, however. Although SGML had been adopted The only thing missing was that it didn't run on the internet.
as a standard by the ISO, it was not w e l l defined as a computer Same story.
language. I also got a strong push back f r o m many people w h o I tried to persuade the people at Grif to add the software
insisted that it w o u l d be too slow. I had to explain that the only needed for sending and receiving files over the Internet, so their
reason SGML was slow was the way it had historically been editor could become a Web browser, too. I told them I would give
implemented. Still, I often had to demonstrate the World Wide them the software outright; they w o u l d just have to hook it i n .
Web program reading an H T M L file and putting it on the screen But they said the only way they w o u l d do that was if we could
i n a fraction of a second before people were convinced. get the European Commission to f u n d the development. They
Some people were intrigued, but many never accepted my didn't want to risk taking the time. I was extremely frustrated.
argument. Rather than enter into useless debate, I simply forged There was a growing group of people w h o were excited about the
ahead w i t h H T M L and showed the Web as much as possible. possibilities of the World Wide Web, and here we had the tech-
Robert and I held a few colloquia open to anyone i n our divi- nology for a true hypertext browser/editor mostly developed, and
sions. We also told people about it at coffee. Occasionally, a we couldn't bridge the gap. Getting Commission funding would
group of people getting ready to do an experiment w o u l d call to have put eighteen months into the loop immediately. This mind-
l a y they were discussing their documentation system, and ask if I set, I thought, was disappointingly different f r o m the more Amer-
could come over and give them my thoughts about it. I ' d meet a ican entrepreneurial attitude of developing something i n the
group of maybe twenty and show them the Web, and perhaps garage for f u n and worrying about funding it when it worked!
they w o u l d n ' t use it then, but the next time through they'd know
In March 1991, I released the WorldWideWeb program to a
about it and a new server w o u l d quietly come into being.
limited number of CERN people w h o had NeXT computers. This
Meanwhile, Robert and I kept putting information on the would at least allow them to w r i t e their own hypertext and make
info.cern.ch server, constantly upgrading the basic guide to new- the Web information that Robert and I were putting on
comers on how to get onto the Web, w i t h specifications and info.cern.ch available to them.
pointers to available software. Word spread w i t h i n the high-energy physics community,
I continued to try to get other organizations to turn their furthered by the cross-pollinating influence of travel. I n M a y
hypertext systems into Web clients. I found out about a powerful 1991 Paul Kunz arrived for a visit f r o m the Stanford Linear Accel-
SGML tool called Grif, developed by a research group at the erator (SLACj i n Palo Alto. Like me, he was an early NeXT enthu-
French lab I N R I A , w h i c h ran on U n i x machines and PCs. A com- siast, and he had come to CERN to w o r k on some common NeXT
pany by the same name, Grif, had since been spun off i n nearby Programs. Since he had the right computer, he was in a position
Grenoble, and I was hopeful its leaders w o u l d entertain the idea to use the Web directly, and he loved it.
of developing a Web browser that could also edit. They had a When Paul returned to SLAC he shared the Web w i t h Louise
beautiful and sophisticated hypertext editor; it w o u l d do graph- Addis, the librarian w h o oversaw all the material produced by
ics, it w o u l d do text in multiple fonts, it w o u l d display the SGML SLAC. She saw it as a godsend for their rather sophisticated but
structure and the formatted document i n two separate windows, mainframe-bound library system, and a way to make SLAC's
and allow changes to be made in either. It was a perfect match. substantial internal catalogue of online documents available to

44
43
w e a v i n g t h e w e b p r o t o c o l s


physicists worldwide. Louise persuaded a colleague who deveU be gan to get e-mail from people who tried to install the software.
oped tools for her to write the appropriate program, and unde would give me bug reports, and "wouldn't it be nice if . . ."
Louise's encouragement SLAC started the first Web server out eports. And there would be the occasional "Hey, I've just set up
side of Europe. server, and it's dead cool. Here's the address."
Seeing that the high-energy physics people at SLAC were ( S VVith each new message I would enter in info.cern.ch a
enthusiastic about the Web, we got more aggressive about p r 0 hypertext link to the new web site, so others visiting the C E R N
moting it within C E R N . In May, Mike Sendall got us an appear- site could link to that address as well. From then on, -interested
ance before the C5 committee, which was continually looking i a people on the Internet provided the feedback, stimulation, ideas,
computing and communications, to explain how useful the Web source-code contributions, and moral support that would have
could be, so management would continue to justify the work. been hard to find locally. The people of the Internet built the
Robert and I wrote a paper, too, "Hypertext at C E R N , " which Web, in true grassroots fashion.
tried to demonstrate the importance of what we were doing. For several months it was mainly the hypertext community
What we hoped for was that someone would say, "Wow! This that was picking up the Web, and the NeXT community because
is going to be the cornerstone of high-energy physics communica- they were interested in software that worked on the platform. As
tions! It will bind the entire community together in the next ten time went on, enough online people agreed there should be a
years. Here are four programmers to work on the project and newsgroup to share information about the Web, so we started one
here's your liaison with Management Information Systems. Any- named comp.infosystems.www. Unlike alt.hypertext, this was a
thing else you need, you just tell us." But it didn't happen. mainstream newsgroup, created after a global vote of approval.
In June we held talks and demonstrations within C E R N , and Another small but effective step to increase the Web's expo-
wrote about the Web in the C E R N newsletter. Because I still had sure was taken when I opened a public telnet server on
no more staff, it was taking longer than I had hoped to get the info.cern.ch. Telnet was an existing protocol, also running over
functionality of the NeXT version onto PCs and Macs and Unix the Internet, that allowed someone using one computer to open
machines. up an interactive command-line session on another computer.
I was still hoping that by spreading the word we could attract Anyone who used a telnet program to log into info.cern.ch would
the attention of more programmers. Since those programmers be connected directly to the line-mode browser. This approach
were unlikely to be high-energy physicists, in August I released had the disadvantage that the user would see the Web as a text-
three thingsâ€"the WorldWiâeWeb for NeXT, the line-mode only read-only system. But it opened the Web to millions of people
browser, and the basic server for any machineâ€"outside C E R N by who could not install a Web browser on their own machine. It
making them all available on the Internet. I posted a notice on meant that someone putting up a Web server could say to "telnet
several Internet newsgroups, chief among them alt.hypertext, to info.cern.ch then type 'go www.foobar.com,'" which was a
which was for hypertext enthusiasts. Unfortunately, there was whole lot easier than requiring them to install a Web browser.
still not much a user could see unless he had a NeXT. The initial home page seen by users of this public service would
Putting the Web out on alt.hypertext was a watershed event. include links to instructions for downloading their own browser.
It exposed the Web to a very critical academic community. 1 Years later we would have to close down the service, since the

46 47
 w e a v i n g t h e w e b p r o t o c o l s


machine couldn't support the load, but by then it would hav e This was a pile of work, but it opened up new possibilities
done its job. a nd also allowed a certain cleaning up as I went along. Jean-
The most valuable thing happening was that people who savvl François arrived at just the right time. For weeks we sat back-to-
the Web, and realized the sense of unbound opportunity, began back in my office spewing out code, negotiating the interfaces
installing the server and posting information. Then they added between each other's modules in remarks over our shoulders.
links to related sites that they found were complementary, or "Can you give me a method to find the last element?"
simply interesting. The Web began to be picked up by people ' "Okay Call it 'lastElement'?"
around the world. The messages from systems managers began to "Fine. Parameters?" ;
stream in: "Hey, I thought you'd be interested. I just put up a
"List, element type. You got it."
Web server."
"Thanks!"
We rolled out the Web-specific code and also had to duplicate
Nicola had to leave the effort in August 1991, since her intern- some of the tools from the NeXTStep tool kit. The result, since a
ship ended and she had to return to college. True to form, Ben collection of bits of code for general use is called a library, we
Segal found yet another gem to replace her. Jean-François Groff called "libwww."
was full of enthusiasm for the whole idea of the Web, and for Unfortunately, C E R N ' s policy with coopérants like Jean-
NeXT. He came to C E R N from France through a "coopérant" pro- François was that they had to leave when their time was up.
gram that allowed the brightest young people, instead of spend- They saw a danger in the staff abusing the program as a recruit-
ing a year in military service, to work for eighteen months at a ment stream, and forbade the employment of any of these people
foreign organization as a volunteer. in any way in the future. When Jean-François came to the end of
By this time we had reached another awkward decision his term, we tried everything we could to allow him to continue
point about the code. Much of the code on the N e X T was in to work on the Web, but it was quite impossible. He left and
the language objective-C. I wanted people to use it widely, but started a company in Geneva, infodesign.ch, probably the very
objective-C compilers were rare. The common language for first Web design consultancy.
portable code was still C , so if I wanted to make it possible for Meanwhile, I had begun to keep logs of the number of times
more people around the Internet to develop Web software, it pages on the first Web server, info.cern.ch at C E R N , were
made sense to convert to C . Should I now, in the interest of prac- accessed. In July and August 1991 there were from ten'to one
tical expediency, convert all my objective-C code back into the hundred "hits" (pages viewed) a day.
less powerful C , or should I keep to the most powerful develop- This was slow progress, but encouraging. I've compared the
ment platform I had? effort to launch the Web with that required to launch a bobsled:
The deciding factor was that Nicola's line-mode browser was Everyone has to push hard for a seemingly long time, but sooner or
written in C . I decided to make the sacrifice and, while keeping later the sled is off on its own momentum and everyone jumps in.
the object-oriented style of my design, downgraded all the com- In October we installed "gateways" to two popular Internet
mon code that I could export from WorldWideWeb on the NeXT services. A gateway was a little program, like that opening up
into the more common C language. Bernd's mainframe server, that made another world available as
48 49
w e a v i n g t h e w e b p r o t o c o l s


part of the Web. One gateway went to the online help system for Web, and they agreed to let us use their dial-in service so we
Digital's VAX/VMS operating system. Another was to Brewster could call the computer back at C E R N .
Khale's WAIS for databases. This was all done to add incentive The next challenge was to get the Swiss modem we had
for any particular individual to install a browser. V M S targeted brought to work with the American electrical system. We bought
the physics community, and WAIS the Internet community. I also a power adapter that would take 110 volts (rather than the Swiss
started an online mailing list, www-talk@info.cern.ch, for techni- 220 volts). Of course it didn't have the right little plug to connect
cal discussions as a forum for the growing community. to the modem. We had to take the modem apart, borrow a solder-
Always trying to balance the effort we put into getting ing gun from the hotel (Robert was rightly proud of this feat!),
involvement from different groups, Robert and I decided we now and wire it up directly. Robert got everything connected, and it
had to promote the Web hard within the hypertext community. A worked.
big conference, Hypertext '91, was coming up in December in We didn't have real Internet connectivity, just a dial-in Unix
San Antonio. Most of the important people in the held would be login, so we could show only the graphic World Wide Web pro-
there, including Doug Ehgelbart, who had created the mouse and gram working on local data. Nonetheless, we could demonstrate
a collaborative hypertext system way back in the 1960s. Though the line-mode browser working live. We were the only people at the
it was difficult to find the time, we cobbled together a paper for entire conference doing any kind of connectivity. The wall of the
it, but didn't do a very good job. It was rejectedâ€"in part because it demo room held project titles above each booth, and only one of
wasn't finished, and didn't make enough references to work in them had any reference to the World Wide Webâ€"ours.
the field. At least one of the reviewers, too, felt that the proposed At the same conference two years later, on the equivalent
system violated the architectural principles that hypertext sys- wall, every project on display would have something to do with
tems had worked on up till then. the Web.
We were able to convince the conference planners to let us
set up a demonstration, however. Robert and I flew to San Antonio
with my N e X T computer and a modem. We couldn't get direct
Internet access in the hotel. In fact, the hypertext community
was so separated from the Internet community that we couldn't
get any kind of connectivity at all. How could we demonstrate
the Web if we couldn't dial up info.cern.ch? Robert found a way.
He persuaded the hotel manager to string a phone line into the
hall alongside the main meeting room. That would allow us to
hook up the modem. Now we needed Internet access. During our
cab ride from the airport, Robert had asked the driver what the
nearest university was and found out that it was the University
of Texas in San Antonio. So Robert called the school and found
some people who understood about the Internet and maybe the

50 5i
 C H A P T E R 5




G o i n g G l o b a l




A s the Web slowly spread around the world, I started to be con-
cerned that people who were putting up servers would not use
HTTP, H T M L , and URIs in a consistent way. If they didn't, they
might unintentionally introduced roadblocks that would render
links impotent.
After I returned to C E R N from San Antonio, I wrote several
more Web pages about the Web's specifications. I would update
them when good ideas came back from other users on the www-
talk mailing list. While this was a start, I wanted to open the Web
technology to wider review. Since everything to date had taken
place on the Internet, and much of it involved Internet protocols,
I felt that the place to get a process going was the Internet Engi-
neering Task Force (IETF), an international forum of people who
chiefly corresponded over e-mailing lists, but who also met physi-
cally three times a year. The I E T F operates on a great principle of
°pen participation. Anyone who is interested in any working
group can contribute.

53
w e a v i n g t h e w e b
g o i n g g l o b a l

As a good software engineer, I wanted to standardize sepa- t 0 C E R N , both Larry and Karen called independently with offers
rately each of the three specifications central to the Web: the UR] to come visit them if I did take leave. I could join Karen as a vis-
addressing scheme, the H T T P protocol by which computers iting researcher at MIT, and join Larry as a visitor at Xerox PARC.
talked to each other, and the H T M L format for hypertext docu- Both invitations were appealing, because both institutions l' '!
ments. The most fundamental of these was the U R I spec. were highly respected and either could give me a much-needed '1 ' 1



The next meeting of the I E T F was in March 1992 in San view of what was happening in the United States rather than !'
Diego, and I went to see how things worked, and how to start a Europe, and in information technology rather than physics. • |
working group. The answer came from Joyce Reynolds, who Encouraged by the enthusiasm of people like Larry and
oversaw one area within the IETF. She said I had to first hold a Karen, Robert and I released notes about the Web on more Inter- " \
"birds-of-a-feather" session to discuss whether there should be a net newsgroups. But we were frustrated by the fact that the
working group. If there was consensus, people at the session Web's use within C E R N itself was very low. We trod a fine line j
could draw up a charter for a working group to begin at the next between dedicating our time to supporting users within C E R N at ,
I E T F meeting. The working group could edit a specification and the risk of neglecting the outside world, and pursuing the goal of > i
take it through to a standard. The subsequent meeting would be
global interactivity at the risk of being bawled out for not stick- 11


held in July in Boston.
ing to C E R N business.
I E T F meetings were characterized by people in T-shirts and By now the Web consisted of a small number of servers, with ' j
jeans, and at times no footwear. They would meet in different info.cern.ch the most interconnected with the rest. It carried a „ j
small rooms and talk excitedly. The networking, of course, was list of servers, which to a degree could coordinate people who | ij
paramount. Compared to Geneva in March, it was a pleasure for were putting information on the Web. When the list became I Ij 1


me to sit with folks outdoors in sunny, warm San Diego. larger, it needed to be organized, so I arranged it in two lists, by " 1



One day over coffee I was seated at a white metal table out in geography and by subject matter. As more servers arrived, it was ( j '
the open air, chatting with Larry Massinter from Xerox PARC and exciting to see how the subjects filled out. Arthur Secret, another ;, I
Karen Sollins, who had been a student of Dave Clark, the profes- student, joined me for a time and set up the lists into what we ( ' |
sor at M I T ' s Laboratory for Computer Science who was very called the Virtual Library, with a tree structure that allowed people I !|
involved with the design of the T C P protocol that had made pos- to find things. |t | 11


sible practical use of the Internet. Karen had stayed on at MIT to
Part of the reason the Web was not being used much within 11


pursue a project called the Infomesh, to create ways computers
CERNâ€"or spreading faster outside C E R N , for that matterâ€"was
could exchange hints to each other about where to find docu-
the lack of point-and-click clients (browsers) for anything other |,
ments they were both interested in.
than the NeXT. At conferences on networking, hypertext, and |'i '|
Larry and Karen asked me what I was doing next. I told them software, Robert and I would point out that for the Web to grow, I ||
I was considering going on sabbatical. I had been at C E R N seven W e
really needed clients for the PC, Macintosh, and Unix. At j ''
years, and while there was no concept of a sabbatical at CERN, I CERN, I was under pressure to make a client for the X Window j '|
felt I needed a break and some new perspective. I needed to s
ystem used by most Unix workstations, but I had no resources. |n !
think about where to take myself and the Web. After I returned
We W e r e s o b u s y trying to keep the Web going that there was no 'i |
54
 w e a v i n g t h e w e b g o i n g g l o b a l


way we could develop browsers ourselves, so we energetically ViolaWWW as a Viola application. This took time and was com-
suggested to everyone everywhere that the creation of browsers licated. But finally, people working on Unix machinesâ€"and
would make useful projects for software students at universities. there were lots of them at corporations and universities around
Our strategy paid off when Robert visited Helsinki University the worldâ€"could access the Web.
of Technology. Several students there decided to make their com- Although browsers were starting to spread, no one working
bined master's project a Web browser. Because the department on them tried to include writing and editing functions. There
was " O T H , " they decided to call the browser Erwise (OTH + seemed to be a perception that creating a browser had a strong
Erwise = "Otherwise"). potential for payback, since it would make information from
By the time it was finished in April 1992, Erwise was quite around the world available to anyone who used*it. Putting as
advanced. It was written for use on a Unix machine running much effort into the collaborative side of the Web didn't seem to
X-Windows. I went to Finland to encourage the students to con- promise that millibnfold multiplier. As soon as developers got
tinue the project after they finished their degrees, and to extend their client working as a browser and released it to the world,
the browser to an editor, but they had remarkably little ongoing very few bothered to continue to develop it as an editor.
enthusiasm for the Web; they had already decided that when they Without a hypertext editor, people would not have the tools
graduated they were going to go on to what they saw as more tan- to really use the Web as an intimate collaborative medium.
talizing or lucrative software projects. No one else around the Browsers would let them find and share information, but they
institute wanted to pick up the project, either. Certainly I couldn't could not work together intuitively Part of the reason, I guessed,
continue it; all the code was documented in Finnish! was that collaboration required much more of a social change in
Another graphical point-and-click browser came at almost the how people worked. And part of it was that editors were more
same time, however. Pei Wei, a very inventive student at U . C . difficult to write.
Berkeley, had created an interpretive computer language called For these reasons, the Web, which I designed to be a medium
Viola, for Unix computers. He had been working on it a long time, for all sorts of information, from the very local to the very global,
and it had powerful functionality for displaying things on the grew decidedly in the direction of the very global, and as a publi-
screen. To demonstrate the power of Viola, Pei decided to write a cation medium but less of a collaboration medium.
Web browser, ViolaWWW. It was quite advanced: It could display There were some pockets of strong internal use. C E R N , even-
H T M L with graphics, do animations, and download small, em- tually, was one. Within Digital Equipment there were a hundred
bedded applications (later known as applets) off the Internet. It was Web servers early on that were not available from the outside.
ahead of its time, and though Pei would be given little credit, Viola- These internal servers were not well publicized, so journalists
WWW set an early standard, and also had many of the attributes could not see them. Years later the media would suddenly "dis-
that would come out several years later in the much-hyped pro- cover" the "rise" of these internal Web networks and invent the
gram Hotjava, which would take the Web community by storm. term intranet, with the notion that they were used largely for
Pei released a test version of his browser on the Web in May internal corporate communications. It seemed somewhat ironic
1992. The only detracting feature was that it was hard for a user to me, since this had been happening all along, and was a prin-
to install on his computer. One had to first install Viola, and then ciple driving the need for the Web in the first place.
56 57
w e a v i n g t h e w e b g o i n g g l o b a l


With Erwise and Viola on board, Robert set out to design a mode browser. Ari, a wild dark Finn took on the server. Each
browser for his favorite computer, the Macintosh. Robert was a ade his mark and put more time and energy into the products
ffl



purist, rather than a pragmatist like me. In the Mac he found the than I could have, in some cases turning them upside down to
realization of his highest ideals of how computers should be: rewrite them into something better. This effort supported a dra-
simple and intuitive to use. But Robert's idealism was sometimes matically growing number of Web sites, and "productized" our
a tough match for the practical need to get a project done. As work so users would find it easy to install and use.
mentioned earlier, I had found a little extra space in the text-editor
code on the N e X T machine, where I could store the U R I address- As the browsers appeared, so did new servers, with ever-increas-
ing information defining each hypertext link. This proved essen- ing frequency. Occasionally, one new server would demonstrate to
tial to being able to make the Web server in a simple way. the community what could be done in a whole new way, and pour
The designers of the Macintosh text editor had a similar fresh energy into the young field. One that impressed me was a
structure, but without the extra space. However, they had set server of information about Rome during the Renaissance. The
aside thirty-two bits for storing the text color, and used only Vatican had lent a (physical) exhibit to America's Library of Con-
twenty-four of them. I suggested we use the spare eight bits, gress. Some of the material in it had been photographed, scanned
and steal a few more from those used for color, which would into a computer, and made available in the form of image files on
not cause any change in the colors that would be noticeable to an FTP Internet server. Then in Europe, Frans van Hoesl, who
users. was aware of the Web, created a hypertext world of this material
Robert was appalledâ€"appalled at the idea of using a field on a Web site. The site took the form of a virtual museum; a
intended for the color for another purpose, appalled at stuffing browser chose a wing to visit, then a corridor, then a room.
the hypertext data into the cracks of the color data. The program On my first visit, I wandered to a music room. There were a
was held up for some time while I tried to persuade Robert that number of thumbnail pictures, and under one was an explanation
taking this admittedly less elegant but simple route would allow of the events that caused the composer Carpentras to present a
him to get on with the rest of the project and actually get the decorated manuscript of his Lamentations of Jeremiah to Pope
Web browser running. I n the end, he accepted my kludge, but in Clement V I I . I clicked, and was glad I had a twenty-one-inch
fact had little time to pursue the program. Later on one summer, color screen: Suddenly it was filled with a beautifully illuminated
Nicola Pellow returned for a few weeks and picked it up, and at score, which I could gaze at probably more easily and in more
one point it was basically working. We named it Samba. detail than I could have done had I gone to the original exhibit at
Every team benefits from a variety of styles, and my collabora- the Library of Congress. This use of the Web to bring distant
tion with Robert was no exception. Robert's insistence on quality people to great resources, and the navigational idiom used to
of presentation would carry us though many papers, demonstra- make the virtual museum, both caught on and inspired many
tions, and presentations. All along, Robert tirelessly trawled for excellent Web sites. It was also a great example of how a combi-
more resources. He ended up getting the students Henrik Frystyk nation of effort from around the world could lead to fantastic things.
Nielsen and Ari Luotonen to join the team. Henrik, an affable Another classic of its time was a server by Steve Putz at
blond Dane, took responsibility for the code library and the line- Xerox PARC. He had a database of geographical information that
58 59
w e a v i n g t h e w e b g o i n g g l o b a l


would generate a virtual map on the fly in response to a user's People at L C S had installed Viola, and M I T was well into the
clicks to zoom and pan. It would prove to be the first of many Vveb. The name "www.mit.edu" was taken very early on by a stu-
map Web servers to come. dent computing club, so "web.mit.edu" would become and remain
Seeing such sites, scientists and government groups, who had the name of MIT's main server. At L C S , I described the ideas
an obligation to make their data available, were realizing it behind the Web to a select group of individuals in the fifth-floor
would be easier to put the information up on the Web than to auditorium. Some of the researchers and administrators wondered
answer repeated requests for it. Typically, when another scien- a bit why I was there. I was trying to see how this creation, which
tist requested their data, they had had to write a custom pro- was really a matter of engineering, fit in from the point of view
gram to translate their information into a format that the person of the research community, what the Web could learn from
could use. Now they could just put it on the Web and ask any- researchers in the field, and why it hadn't happened before.
one who wanted it to go get a browser. And people did. The At the I E T F meeting I held my birds-of-a-feather session to
acceptability of the Web was increasing. The excuses for not investigate forming a working group to standardize the U R I spec,
having a browser were wearing thinner. The bobsled was start- as Joyce Reynolds had suggested. We met in a small room at the
ing to glide. Hyatt Hotel. I presented the idea of a universal document
identifierâ€" my initial name for itâ€"and said I was interested in it
As June 1992 approached, I increasingly felt the need for a sab- being adopted as an Internet standard. A number of things went
batical. David Williams, head of my division at C E R N , had seen less than smoothly. The open discussion was great. I felt very
this coming and was ready with an offer I couldn't refuse. He much in the minority. There was another minority who seemed
explained that I could go away for a year and have my job when I to resent me as an intruding newcomer.
returned. However, during that year I would lose my C E R N Even though I was asking for only a piece of the Web to be
salary and benefits, which were quite good, and I would have to standardized, there was a strong reaction against the "arrogance"
pay all my travel expenses. As an. alternative, David said I could of calling something a universal document identifier. How could I
go away for an extended business trip for three months and he be so presumptuous as to define my creation as "universal"? If I
would pay me a per diem rate for this "extended duty travel," on wanted the U D I addresses to be standardized, then the name
top of my ongoing salary and benefits. Not surprisingly, I chose "uniform document identifiers" would certainly suffice. I sensed
the second option. My wife and I planned a three-month mixture an immediate and strong force among the people there. They
of work and vacation. I would visit MIT's Laboratory for Com- were trying to confine the Web to some kind of tidy box: Nothing
puter Science (LCS) in Cambridge, Massachusetts, and also attend could be universal. Others viewed the I E T F as a place where
the I E T F meeting in neighboring Boston. Then we would vaca- something universal might be created, but that something was
tion in New Hampshire, and end up in the San Francisco area ot going to be the Web. Those tensions would continue through
fl



where I would visit Xerox PARC. that I E T F meeting and subsequent ones. Some people wanted to
The summer turned out to be a great opportunity for me to integrate the Web with other information systems, which directly
take a snapshot of the state of the Web's penetration and accep- begged the point, because the Web was defined to be the integra-
tance in the States. tion of all information systems.
60 61
 w e a v i n g t h e w e b g o i n g g l o b a l



I tried to explain at the session how important it was that the arbitrary decision (like which punctuation characters to use) that
Web be seen as universal, but there was only so much time, and I I had already made, and changing it would only mean that mil-
decided not to waste my breath. I thought, What's in a name? If it lions of Web browsers and existing links would have to be
went through the standards process and these people agreed, and changed. After months of rather uncontrolled arguing in the
all I needed was to call it uniform, as long as I got the right spec IETF, it seemed that they had to take either all of the Web, or
that was fine by me. I was willing to compromise so I could get none of it. In the end I wrote a specification on how URIs were
to the technical details. So universal became uniform, and docu- used on the Web, and issued it to the I E T F community as an
ment became resource. informational "Request for Comment 1630." While hurried and
As it turns out, it had been important to nail down the name, with a few mistakes, it was a foothold for future progress. The
because behind the name was the fundamental philosophical whole affair would also have gone more smoothly had I been
underpinnings of what the Web was trying to be. Ultimately, the more forceful about the points on which I was prepared to nego-
group did decide to form a uniform resource identifier working tiate and those on which I was not.
group. However, they decided that identifier wasn't a good label My stay at L C S had been more inspiring, and the same was
for what the Web used. They wanted to emphasize that people true when I went to Xerox PARC. Being security conscious, PARC
could change the URIs when moving documents, and so they had many experimental servers available internally, protected
should be treated as some sort of transitive address. Locator was behind a firewall built into their system that prevented outsiders
chosen instead, like a branding, a warning mark on the technol- from illegally gaining electronic access. There was a special way of
ogy. I wanted to stick with identifier, because though in practice getting a connection from inside to outside. They were not using
many URIs did change, the object was to make them as persis- Viola because it had to be compiled with special code to make this
tent as possible. We argued, but at the I E T F the universal connection, so the first thing I did on arrival was to do that.
resource identifier became U R L , the uniform resource locator. In I also visited other important actors in the Web world while
years ahead the I E T F community would use the U R L acronym, in the San Francisco area. When going to PARC I would bike in
allowing the use of the term URI for what was either a U R L or every day past SLAC. I stopped in to see Paul Kunz and Louise
something more persistent. I use the general term URI to empha- Addis, early promoters and implementers of the Web. I also got
size the importance of universality, and of the persistence of together with Pei Wei, who was still at U . C . Berkeley. Although
information. Viola was attracting some attention, the difficulty in installing it
Progress in the U R I working group was slow, partly due to limited its appeal. I met Pei at a café outside San Francisco to try
the number of endless philosophical rat holes down which tech- to persuade him to make installation easier, and to give editing
nical conversations would disappear. When years later the URI power to his browser as wellâ€"still my ideal. But Pei's interest
working group had to meet twelve times and still failed to agree was always in Viola as a computer language; he saw the Web as
on a nine-page document, John Klensin, the then I E T F Applica- just one application of it. I tried to encourage but not push. After
tions Area director, was to angrily disband it. Sometimes there
a l
l , Viola was broadening the Web's reach tremendously. Part of
was a core philosophy being argued, and from my point of view
m
y reason to meet him was simply to say, in person, "Thank you,
that was not up for compromise. Sometimes there was a basically well done."

63
w e a v i n g t h e w e b g o i n g g l o b a l


Pei's unassuming demeanor and lack of arrogance about his Different people had tackled different aspects of the social
ideas were remarkable given his product, which was great. When implications of hypertext. For Ted, hypertext was the opposite of
I congratulated him and told him that further development would copyright. The whole idea of Xanadu was driven by his feeling
make Viola the flagship of Web browsers, Pei smiled, but he that anybody should be able to publish information, and if some-
would reserve his program as his own research tool. He would go one wanted to use that information, the creator ought to be auto-
on to join the Digital Media group at O'Reilly Associates in matically recompensed. One of the reasons Xanadu never took
Sebastopol, California, run by Dale Dougherty, one of the early off was Ted's insistence on a pricing mechanism, and the diffi-
Web champions, which was creating various Internet products. • culty of creating one that was consistent across the whole world.
He used Viola to demonstrate what online products could look In theory this would be possible on the Web with certain exten-
like using different styles. sions, and a system of "micropayments"â€"small debentures against
Because the installation process was a little too complex, Viola a person's bank accountâ€"would allow automatic payments in
was destined to be eclipsed by other browsers to come. Indeed, very small quantities. I was not keen on the idea of having only
there was already competition between Web browsers. While one business model for paying for information. But I was keen on
Erwise and ViolaWWW competed as browsers for the X Window meeting Ted.
system on Unix, Tony Johnson at SLAC entered the fray. A physi- We had corresponded only a few times via e-mail, and the
cist, he had developed another browser for X called Midas, partly fledgling relationship we had was a strange one for me at least,
because he liked to see a program written well, and partly because for a long time I owed Ted money. I had first heard of
because in his project he wanted to use the Web to disseminate Ted in 1988 when reading about hypertext. His main book at the
his information, and wanted a browser he could control. He used time was Literary Machines, published by the Mindful, Press,
a nice conceptual model, the programming was very clean, and it which Ted operated as a one-man publishing house. Some time
allowed him, for example, to import images in a very flexible way. later I got around to sending him an order for the book with a
I met Tony in his office at SLAC. Although he gave presenta- check written out in U.S. dollars drawn on my Swiss bank
tions around SLAC about Midas, and used it himself, he was as account. Swiss checks were very international, with a space for
reluctant as Pei or the Erwise group to join in my effort at the amount and a space for the currency type, but I didn't realize
C E R N , even though it would probably provide extra resources. ; American banks didn't accept them. He sent the book, but I
Tony was and is first and foremost a physicist, and he didn't like didn't succeed in paying, since he didn't take credit cards and I
the idea of supporting Midas for a group any wider than that of didn't have U.S. checks.
his colleagues. And so it had stayed. I called him up from PARC and found
The month I was spending in California was coming to a that he lived on a houseboat in Sausalito, across the Golden Gate
close, and soon my family and I would have to return to Geneva. Bridge from San Francisco. It was the place closest to where
But I could not go back without making one more stop, which 1 things were happening that was sufficiently eccentric for him to

knew would be perhaps the greatest treat of the summer. Ted
llVe
- Xanadu had been picked up by Autodesk, and Ted had some
dignitary position with the company. But the day I was scheduled
Nelson, who had conceived Xanadu twenty-five years earlier,
meet him for lunch was a sad one. That very morning
lived close by, and I had to meet him.
64 65
w e a v i n g t h e w e b


C H A P T E R 6
Autodesk had decided Xanadu was an impractical project after
all. They were dropping it, leaving the project homeless.
Ted kindly bought me an Indian lunch anyway, and then we
went back to his office, which seemed to be an attic in a pyramid
building on the Sausalito shore. It was full of copies of his books.
I gave him the money I owed and he promptly gave me a second
book, autographed. We talked about all manner of things, but not
B r o w s i n g
a lot about Autodesk.
After lunch Ted walked me to my car in the parking lot. I
took out my 35-mm camera from the trunk to capture the
moment. I asked Ted, with some embarrassment, if he would
mind posing for my scrapbook. He replied, "Certainly, not at all.
I understand completely." He then produced from his knapsack a
video camera to shoot some video footage of me. Before he did,
though, he held the camera at arm's length, pointed it at his
head, and shot a little bit of himself explaining that this was Tim
Berners-Lee he would be filming, and what the significance was.
Ted explained to me that it was his objective to lead the most
interesting life he could, and to record as much as possible of By January 1993 the number of known servers was increasing
that life for other people. To which end he amassed a huge num- faster, up to about fifty. The Erwise, Viola, and Midas- browsers
ber of video clips, which were indexed with an image of his own were generally available for use on the X Window system. Samba
head; that way, he could skip through, and whenever he saw his was working, though not complete, for the Mac. But to me it was
head he could listen for a description of the next clip to come. clear there was growing competition among the browsers, even if
it was on a small scale. Many of the people developing browsers
The summer of 1992 had been a thrilling time for me. The Web were students, and they were driven to add features to their ver-
was being seen and used in many more places, and more people sion before someone else added similar features. They held open
were developing browsers for it. I looked over the logs showing discussions about these things on the www-tajk mailing list, pre-
the traffic that the first Web server, info.cern.ch, had been getting serving the open social processes that had characterized Internet
over the last twelve months. The curve showing the number of software development. But there was still an honorable one-,
daily hits was a dramatic exponential, doubling every three to four upmanship, too.
months. After one year, the load had grpwn by a factor of ten. One of the few commercial developers to join the contest was
Dave Raggett at Hewlett-Packard in Bristol, England. He created
a browser called Arena. HP had a convention that an employee
could engage in related, useful, but not official work for 10 percent
66 67
 w e a v i n g t h e w e b b r o w s i n g


of his or her job time. Dave spent his "10 percent time," plus a l o t This was in total contrast to any of the other student develop-
of evenings and weekends, -on Arena. He was convinced that ers Marc was not so much interested in just making the program
hypertext Web pages could be much more exciting, like magazine w o r k as in having his browser used by as many people as possi-
pages rather than textbook pages, and that H T M L could be used ble. The was, of course, what the Web needed.
to position not just text on a page, but pictures, tables, and other The resulting browser was called Mosaic. In February 1993
features. He used Arena to demonstrate all these things, and to NCSA made the first version available over the Web. I tried it at
experiment with different ways of reading and interpreting both CERN. It was easy to download and install, and required very lit-
valid and incorrectly written H T M L pages.
tle learning before I had point-and-click access to the Web.
Meanwhile, the University of Kansas had, independently of the Because of these traits, Mosaic was soon picked up more rapidly
Web, written a hypertext browser, Lynx, that worked with'80 x 24 than the other browsers. Mosaic was much more of a product.
character terminals. More sophisticated than our line-mode It troubled me in a way that NCSA was always talking about
browser, Lynx was a "screen mode" browser, allowing scrolling Mosaic, often with hardly a mention of the World Wide Web. Per-
backward and forward through a document. It had, like Gopher, haps it was just pure enthusiasm.
been designed as a campus-wide information system, and the team I was scheduled to give a presentation to the Fermi National
joked that Lynxes ate Gophers. Lou Montulli, a student, adapted it Accelerator Laboratory (Fermilab) in Chicago in March, which
to the Web and released a Web browser, Lynx 2.0, in March 1993. had put up a server as SLAC had done. I decided I would visit
Developing browsers had become a good vehicle for students NCSA as well, since it was only a few hours' drive away.
and engineers to show off their programming skills. David " While in Chicago I met Tom Bruce, a stage manager turned
Thompson, a manager at the National Center for Supercomputing systems administrator turned programmer who had recently
Applications (NCSA) at the University of Illinois at Urbana- cofounded the Legal Information Institute at Cornell University,
Champaign, wanted students to take a crack at it. He down- to provide online legal information and law findings. He thought
loaded Viola, got it running, and demonstrated its use with the the Web was just what the institute needed to distribute this
C E R N server to the rest of NCSA's Software Design Group. information to the legal community. He had realized that most
Marc Andreessen, a student, and Eric Bina, a staff member, lawyers used I B M PCs or compatibles, which ran the Windows
decided to create a browser for X. Eric was somewhat like Pel operating system, and would need a browser. So he had written
Wei, quietly programming the H T M L code and making the thing Cello, a point-and-click browser for Windows. It was at alpha
work. Marc maintained a near-constant presence on the news- release (an early test version) in March, and he had come to
groups discussing the Web, listening for features people were ask- Chicago to give a talk to the legal community about it. For the
ing for, what would make browsers easier to use. He would first time, people could see the Web in its multicolor, multifont
program these into the nascent browser and keep publishing new glory on the world's most widespread computing platform.
releases so others could try it. He listened intently to critiques, I found Tom in an auditorium just after he had finished his
almost as if he were attending to "customer relations." Nourished, talk. His laptop computer was still on, with its screen projected
it was said, by large quantities of espresso, he would fix bugs and onto a big movie screen at the head of the room. There he
add little features late at night in reaction to user feedback. demonstrated Cello to me, the two of us sitting alone in this big
68 '69
w e a v i n g t h e w e b b r o w s i n g


room looking up at this big image of the Web. He had multiple ance between competing demands on developers' time. But it
fonts, colors, and user-selectable styles. He used á dotted line was also true that most were more excited about putting fancy
around text denoting a hypertext link, which fit with Windows display features into the browsersâ€"multimedia, different colors
conventions. I found out in talking with him afterward that he and fontsâ€"which took much less work and created much more
had worked professionally with lighting and audiovisual equip- buzz among users. And Marc, more than anyone, appeared inter-
ment in the theater. I had done the same thing in an amateur ested in responding to users' wants.
way. We shared an enthusiasm for the vocation, and hit it off. I sensed other tensions as well. There was a huge difference
I asked Tom, and Ruth Pordes, my host at Fermilab and a source in style among the three men, and each seemed to be thinking
of honest wisdom, to come with me to meet Marc Andreessen separately rather than as a team. Eric, the staffer, was quiet.
and the folks at NCSA. Ruth drove us down across the seemingly Marc, the student, gave the appearance that he thought of this
interminable cornfields. As someone who had been living in meeting as a poker game. Hardin was very academic, the con-
Geneva, I was struck by a remarkable lack of mountains. summate professor in a tweed jacket. He was interested in the
The three of us found the Software Development Group, social implications of the Web as well as the technology, and in
though it was not in the imposing brick and green-glass buildings sociological studies of the Web. For him Mosaic was a sequel to a
that housed most of NCSA, but in an annex to the oil-chemistry project NCSA already had, a multimedia hypertext system called
building. We met Eric, Marc, and the group's leader, Joseph Collage.
Hardin, in a basement meeting room. To add to my consternation, the NCSA public-relations depart-
All my earlier meetings with browser developers had been ment was also pushing Mosaic. It wasn't long before the New York
meetings of minds, with a pooling of enthusiasm. But this meet- Times ran an article picturing Hardin and Larry Smarr, the head of
ing had a strange tension to it. It was becoming clear to me in
1
NCSA, (not Marc and Eric!) sitting side by side at terminals run-
the days before I went to Chicago that the people at NCSA were ning the Mosaic browser. Once again, the focus was on Mosaic, as
attempting to portray themselves as the center of Web develop- if it were the Web. There was little mention of other browsers, or
ment, and to basically rename the Web as Mosaic. At NCSA, even the rest of the world's effort to create servers. The media,
something wasn't "on the Web," it was "on Mosaic." Marc which didn't take the time to investigate deeper, started to portray
seemed to sense my discomfort at this. Mosaic as if it were equivalent to the Web.
I dismissed this as a subject of conversation, however, and I returned to C E R N uneasy about the decidedly peremptory
made my now-standard case for making the Mosaic browser an undertones behind NCSA's promotion of Mosaic. NCSA quickly
editor, too. Marc and Eric explained that they had looked at that started other projects to get Mosaic onto PCs running Windows,
option and concluded that it was just impossible. It couldn't be and onto Macintoshes.
done. This was news to me, since I had already done it with the
World Wide Web on the NeXTâ€"though admittedly for a simpler The rise of different browsers made me think once again about
version of H T M L . standardization. The I E T F route didn't seem to be working. I
Still, I was amazed by this near universal disdain for creating thought that perhaps a different model would. I got more enthused
an editor. Maybe it was too daunting. Or maybe it was just a bal- about the idea during a seminar at Newcastle University in my

70 71
 w e a v i n g t h e we.b b r o w s i n g


native England, organized by International Computers Ltd. The This was an act of treason in the academic community and
spring weather was wet and dark. We were bused through the rainy the Internet community. Even if the university never charged any-
evening from the seminar to dinner. On the way back I sat next to one a dime, the fact that the school had announced it was reserv-
David Gifford, who happened to be a professor at MIT's LCS. I told ing the right to charge people for the use of the gopher protocols
him I was thinking of setting up some kind of body to oversee the meant it had crossed the line. To use the technology was too risky.
evolution of the Web. I wondered what kind of structure might Industry dropped gopher like a hot potato. Developers knew
work, and where to base it. He said I should talk to Michael they couldn't do anything that could possibly be said to be
Dertouzos about it. He explained that Michael was the director of related to the gopher protocol without asking all their lawyers
1




LCS, and said he thought Michael might be interested in doing first about negotiating rights. Even if a company wrote its own
something. I expressed happy surprise, noted "mld@hq.lcs.mit.edu," gopher client or, server, the university could later sue for infringe-
and promptly e-mailed him when I got back to C E R N . ment of some intellectual property right. It was considered dan-
I was further motivated by another Internet phenomenon gerous as an engineer to have even read the specification or seen
that had recently taken place. The gopher information system at any of the code, because anything that person did in the future
the University of Minnesota had started at about the same time could possibly be said to have been in some way inspired by the
as the Web. It was originally created as an online help system for private gopher technology.
the university's computing department and spread to become a At the March 1993 I E T F meeting in Columbus, Ohio, held
campuswide information system that also allowed people to after the announcement, I was accosted in the corridors: "Okay,
share documents over the Internet. Instead of using hypertext this is what happened to gopher. Is C E R N going to do the same
and links, it presented users with menus, taking them eventually thing with the WWW?" I listened carefully to peoples' concerns
to documents normally in plain text. I had found that some people, and to what they said they would or would not find acceptable. I
when they saw the Web, thought hypertext was confusing, or also sweated anxiously behind my calm exterior.
worried that somehow they would get lost in hyperspace when During the preceding year I had been trying to get C E R N to
following a link. Of course, this could happen in gopherspace release the intellectual property rights to the Web code under the
too, but computer users were familiar with menus, so the pro- General Public License (GPL) so that others could use it. The
gram didn't seem as foreign. GPL was developed by Richard Stallman for his Free Software
It was just about this time, spring 1993, that the University Foundation, and while it allowed things to be distributed and
of Minnesota decided it would ask for a license fee from certain used freely, there were strings attached, such that any modifica-
classes of users who wanted to use gopher. Since the gopher tions also had to be released under the same G P L . In the fallout
software was being picked up so widely, the university was of the gopher debacle, there were already rumors that large com-
.going to charge an annual fee. The browser, and the act of panies like I B M would not allow the Web on the premises if
browsing, would be free, and the server software would remain there was any kind of licensing issue, because that would be too
free to nonprofit and educational institutions. But any other constraining. And that included the G P L .
users, notably companies, would have to pay to use gopher server C E R N had hot yet made up its mind. I returned from Columbus
software. and swiftly switched my request, from getting a G P L to having
72 73
w e a v i n g t h e w e b


the Web technology put in the general public domain, with no C H A P T E R - 7

strings attached.
On April 30 Robert and I received a declaration, with a
C E R N stamp, signed by one of the directors, saying that C E R N
agreed to allow anybody to use the Web protocol and code free of
charge, to create a server or a browser, to give it away or sell it,
without any royalty or other constraint. Whew! C h a n g e s




M y experience at NCSA, and the near disaster over licensing,
made me more convinced than ever that some kind of body was
needed to oversee the Web's development. The Web's fast growth
added to my feeling. The Web was starting to change phase. Some
people were still sending me e-mail about putting up new servers.
But others were not; they just started them. C E R N and I were
beginning to blend into the background hum. Web activity was
increasing at a relentlessly steady, exponential rate. It being mid-
summer, I once again graphed the number of people who were
accessing the C E R N server, info.cern.ch. It was now taking ten
thousand hits a day. The rate was incredible, still doubling every
three or four months, growing by a factor of ten every year, from
°ne hundred hits a day in the summer of 1991, to one thousand in
the summer of 1992, to ten thousand in the summer of 1993.
1 no longer had to push the bobsled. It was time to jump in
and steer.

74 75
w e a v i n g t h e w e b c h a n g e s


I did not want to form a standards body per se, but some dows, and the Mac. Dale was wondering himself where the Web
kind of organization that could help developers of servers and was going, and felt he could find out, and perhaps also help people
browsers reach consensus on how the Web should operate. With make it go somewhat sensibly, by getting everyone together.
Mosaic picking up the ball and running single-handedly for the About twenty-five of the early Web developers gathered at
goal line, and more and more gopher users considering the Web, O'Reilly's offices in Cambridge. There was Lou Montulli, who
evidence was mounting that "the Web" could splinter into vari- had adapted Lynx for the Web, and his boss; a group from NCSA
ous factions â€"some commercial, some academic; some free, including Eric Bina, Marc Andreessen, Chris, Wilson, who was
some not. This would defeat the very purpose of the Web: to be porting Mosaic to the PC, and Alex Totic, who was porting it to
a single, universal, accessible hypertext medium for sharing the Mac; Tom Bruce, author of Cello; Steve Putz from Xerox
information. PARC, of map server fame; Pei Wei, author of Viola; and others.
I talked to people at C E R N about starting some kind of con- The focus of the meeting was on defining the most important
sortium. I also swapped e-mails with Michael Dertouzos at MIT's things to do next for the Web development community. I n his
Laboratory for Computer Science. Michael seemed very receptive friendly, encouraging way, Dale got us all talking. I brought up
to the idea. A frequent visitor to Europe and his native Greece, the general idea for a Web consortium. We discussed what it
he arranged to meet me in Zurich on February 1, 1994. could be like, whether it should be a consortium or an organiza-
I took the train from Geneva to Zurich not knowing quite tion or a club. At one point I put the words Club Web up on the
what Michael wanted, nor what I did. We met at a pleasant café whiteboard. . . . Well, it was an option. I led a brainstorming ses-
in the old town, and over some characteristic Zurich-style veal sion to list the needs for the next few months, covering the walls
and Rósti, we ended up sketching plans for the top levels of a on all sides with ideas grouped to make some kind of sense.
consortium. We both returned to our homes to mull over our The event was quite a bonding occasion for some members of
ideas. the community. Even for hard-core devotees of the Internet it's
It seemed more than a bit serendipitous that the first WWW fun to meet face-to-face someone you have communicated with
Wizards Workshop was scheduled to be held only a month or so only by e-mail. During the meeting several people commented on
later . . . in Cambridge, Massachusetts, just a few blocks from how surprised they were that Marc, who had been so vocal on
MIT. It had been set up by Dale Dougherty of O'Reilly Associ- the Internet, was so quiet in person. A few of us were taking
ates, who again quietly managed to gather the flock. photos, and Marc was the only one who basically refused to be
O'Reilly had just published E d Krol's book Whole Earth Inter- photographed. I managed to sneak a picture of him with a tele-
net Catalog, which was really the first book that made all this photo lens, but for all his physical size and lack of hesitation to
Internet stuff accessible to the public. When I had proofread it, come out blaring on the www-talk newsgroup, he and the others
on the train in Chicago going to meet Tom Bruce, the World Wide from NCSA were remarkably self-conscious and quiet.
Web occupied just one chapter; the rest was about how to use all I returned to C E R N with a clearer vision that a consortium
the various Internet protocols such as F T P and telnet and so on. was needed. Then one day the phone in my office rang. It was
But the traffic on the Web was increasing fast, and NCSA had just reception saying there were four people from Digital Equipment
released working versions of the Mosaic browser for Unix, Win- Corporation to see me. Now, C E R N was not a place where people

76 77
w e a v i n g t h e w e b
c h a n g e s

just turned up at reception. It is international, it's huge, people By October there were more than two hundred known H T T P
have to come from a long way, they need an escort to hnd their servers, and certainly a lot more hidden ones. The European I;
way around. But suddenly this group of people in suits was here. Commission, the Fraunhofer Gesellschaft, and C E R N started the
I quickly commandeered an available conference room. There first Web-based project of the European Union, called Webcore, '
were three men and one woman: Alan Kotok, the senior consul- for disseminating technological information throughout the for- \j
tant; Steve Fink, a marketing man; Brian Reed, D E C ' s Internet mer Soviet bloc countries in Europe. Then in December the ,
guru at the time; and Gail Grant, from the company's Silicon Val- media became aware, with articles in major publications about j
ley operations. the Web and Mosaic, and everything was being run together. h
Alan had been pushing D E C in the direction of the Web Meanwhile, the community of developers was growing. It ¡1
ever since he had been shown a Web browser, and management would be obviously exciting to hold a World Wide Web confer- !'
had asked Steve to put together a team to assess the future of ence to bring them together on a larger scale than the Wizards |"
the Internet for D E C . Steve explained that they would be Workshop had done. I had already talked to Robert about it, and
largely redesigning D E C as a result of the Web. While they saw now the need was more pressing. He got the go-ahead from 1



this as a huge opportunity, they were concerned about where C E R N management to organize the first International WWW ^
the Web was headed, worried that the Web was perhaps defined Conference and hold it at C E R N . Robert was excited and checked ,|
by nothing more than specifications stored on some disk sitting the schedule of availability for the auditorium and three meeting . i
around somewhere at C E R N . They wanted to know what rooms. There were only two dates open within the next several j*
C E R N ' s attitude was about the future path of the Web, and months. He booked one of them immediately. He came back and 1
*
whether they could rest assured that it would remain stable yet said, "You don't have to do anything. I'll do everything. But this 1




evolve. is the date it has to be held." 1
'
I asked them what their requirements were, what they felt I said, "Well, Robert, that's fine, except that it's the date that '
was important. They felt strongly that there should be a neutral my wife and I are expecting our second child." He realized j,
body acting as convéner. They were not interested in taking over there were things that could be moved and things that couldn't ,
the Web, or having some proprietary control of it. But they really be. He sighed and went back to see if the other date was still j ,
wanted a body of oversight to which they could become attached. available. It was, but the date, at the end of May, was earlier I
They wondered if C E R N would do this. than the first one, and it left us with short notice to get it all
For me this was a listening meeting. It was important input together. <
into the decision about what to do next. I told them I had talked Robert went about quickly coordinating all the bits and
with M I T about perhaps running a group. It might be modeled Pieces needed for a conference, including speakers. One of the '
after the X Consortium, which M I T had organized to take Bob hrst people he called was Joseph Hardin at NCSA. But Hardin's !;
Scheifler's X Window system from his initial design to a platform response to Robert was: "Oh, well, we were thinking of holding a
used by almost all Unix workstations. It seemed to strike them as inference, and May is basically when we were going to do it, in
an exceptional idea. Chicago. Would you mind canceling your conference so we can
8° ahead with ours?"

78 79
w e a v i n g t h e w e b


Robert debated with himself for only a moment. There was the Internet via a local telephone call-. They provided all the soft-
honor and pride at stake here, but also the future direction of the ware a subscriber required. This made Internet in a Box
Web. The conference was the way to tell everyone that no one unneeded. And it was a strong indicator of the rapid commercial-
should control it, and that a consortium could help parties agree ization of "the Net."
on how to work together while also actually withstanding any A short month later Navisoft Inc. released a browser/editor
effort by any institution or company to "control" things. Feeling for the P C and the Mac, which was remarkably reminiscent of
that perhaps NCSA was again trying to beat us to the punch, my original World Wide Web client. Navipress, as it was called,
Robert told Hardin, "Well, if you had planned your conference so allowed a person to browse documents and edit them at the same
long ago then you certainly would have told us about it by now. time. There was no need to download something explicitly, edit it
So, sorry, we intend to go forward with ours." He pointed out that with a different mode, then upload it againâ€"finally, a browser
we had already booked the space and had passed the point of no that also functioned as an editor. I was very glad to hear of it.
return. NCSA decided to hold a second WWW conference in Usually when we had talked about the principles of the Web,
Chicago in November. most people just didn't get it. But Dave Long and the people at
As 1994 unfolded, more signs emerged that the general public Navisoft had gotten it, miraculously, just by reading everything
was beginning to embrace the Web. Merit Inc., which ran the we had written on info.cern.ch and by following the discussions
Internet backbone for the National Science Foundation, mea- of the Web community. Navipress was a true browser and editor,
sured the relative use of different protocols over the Internet. In which produced clean H T M L .
March. 1993, Web connections had accounted for 0.1 percent of I talked again with Michael Dertouzos about forming a con-
Internet traffic. This had risen to 1.percent by September, and 2.5 sortium. In February he invited me to MIT's L C S to see if we
percent by December. Such growth was unprecedented in Inter- could work out details we'd both be happy with. He took me to
net circles: lunch at the Hyatt, which I understood was his usual place for
In January, O'Reilly had announced a product dubbed "Inter- serious discussion. The doorman knew him so well he had a
net in a Box," which would bring the Internet and Web into cordoned-off space waiting for Michael's B M W at any time.
homes. It was already possible for anyone to download, free, all Michael had helped put together other high-level organizations
the browsers, TCP/IP, and software needed to get on the Internet that included academic, industry, and government people, and
and Web, but a user had to know a lot about how to configure was assuming that a similar model would hold for a Web consor-
them and make them work together, which was complicated. tium. But when he asked me where I wanted such an organiza-
Neither the Internet nor the Web had initially been set up for tion to reside, I hesitantly mentioned I didn't want it to be based
home or individual business use; they were meant for universi- just af MIT: I wanted it to be international. I didn't want to defect
ties, researchers, and larger organizations. O'Reilly's product put from Europe to the States. I thought there should be a base in
it all together. . All a user had to do was install it on his computer, Europe and a base in the States.
and pay phone charges for his connection to the Internet. To my delight, this made perfect sense to Michael. He was
Soon thereafter, however, many Internet service providers happy to have LCS be part of what he called a two-legged beast. Of
started to spring upâ€"local companies that would give access to Greek descent, Michael had made many transatlantic connections

8o 8i
w e a v i n g t h e w e b c h a n g e s


over the years, and had always been interested in fostering joint activity. Leveraging Marc's skills, NCSA pushed Mosaic hard,
efforts between the Old World and the New. I had hit not a snag, from being a great idea seen in Viola to a must-have product that
but one of Michael's hot buttons. We returned to L C S with joint was going to be on every desktop. Andreessen and Clark set out
enthusiasm and warmth. aggressively to conquer the entire market. To do so they used an
Michael later introduced me to his associate director, Al Vezza, unprecedented marketing policy: They released their product for
who had helped Bob Scheifler set up the X Consortium and run it free, so it would be picked up widely and quickly; all someone
from LCS for years. Al took me into his office and asked me blunt had to do was download it from the Internet. They also seemed
questions about the business end of a consortium, questions to to follow the unprecedented financial policy of not having a busi-

which I had no answers, questions about the organization struc- ness plan at first: they decided not to bother to figure out what

ture and the business model. Fortunately, A l had answers. He had the plan would be until the product was world famous and

set up these kinds of things for the X Consortium, and was happy omnipotent.

to'do the same again. The X Consortium plan had been so well The arrival of Web software and services as a commercial
defined that A l ended up convincing me to follow a similar model. product was a very important step for the Web. Many people
C E R N clearly had first option to be the European host. Michael, would not really want to use the Web unless they could be sure
Al, and I had pretty much assumed that C E R N would sign on. I they could buy the products they needed from a company with

returned to Geneva and began a series of talks about C E R N all the usual divisions, including customer support. Robert and I

assuming this new role. had spent so much time trying to persuade companies to take on

As the talks ensued, Marc Andreessen, who had left NCSA to the Web as a product. At last, it had happened.

join Enterprise Integration Technology (EIT), had met business- People began to ask me whether I was planning to start a

man Jim Clark. Together they founded Mosaic Communications company. Behind that question, maybe they were wondering if I •

Corp. The two rapidly hired Lou Montulli of Lynx fame, hired felt the rug had been swept out from beneath my feet by Marc

away the core Mosaic development team from NCSA, and set out Andreessen and Jim Clark. Of course, I had several options apart
from starting a consortium. I had actually thought about starting
to commercialize their browser. They'd soon relocate to Moun-
a company with the working name Websoft, to do much the
tain View, California, and in April 1994 would rename them-
same as Netscape. (The name was later taken by a real company.)
selves Netscape.
But at this point, starting a company was by no means a guaran-
Despite the news articles hailing it as the first step of an
tee of future riches. It was a financial risk like any startup, and a
Internet revolution, Netscape's start was very natural. The
considerable one in this case, since there was not even a clear
Mosaic team, unlike any of the other browser teams, had always
market yet.
operated much more like a product development team than a
research team. They were much more aware of Mosaic's brand- Furthermore, my primary mission was to make sure that the
ing, of customer relations, marketing, and delivery. NCSA delib- Web I had created continued to evolve. There were still many
erately adapted Mosaic for multiple platforms so it would reach a things that could have gone wrong. It could have faded away,
large audience. Unlike C E R N , NCSA never doubted for a been replaced by a different system, have fragmented, or
moment that creating commercial products was an appropriate changed its nature so, that it ceased to exist as a universal

82 83
w e a v i n g t.he w e b c h a n g e s


medium. I remembered what Phil Gross, chairman of the IETF, the industry movement to get Web products into the marketplace
had once said about gopher when it was still rising in popularity: and into people's real lives. I did talk to several companies and
"Things can get picked up quickly on the Internet, but they can visited a few labs to evaluate this possibility, but there didn't
be dropped quickly, too." My motivation was to make sure that seem to be a good match.
the Web became what I'd originally intended it to beâ€"a univer- Starting a consortium, therefore, represented the best way for
sal medium for sharing information. Starting a company would me to see the full span of the Web community as it spread into
not have done much to further this goal, and it would have risked more and more areas. My decision not to turn the Web into my
the prompting of competition, which could have turned the Web own commercial venture was not any great act of altruism or dis-
into a bunch of proprietary products. Theoretically, it would have dain for money, of which I would later be accused.
been possible to have licensed the technology out, but the swift
demise of gopher reasoned against that. While the press was making a big deal about Mosaic Communi-
I also realized that by following the consortium route I could cations, the first World Wide Web conference was now fast
keep a neutral viewpoint, affording me a much clearer picture of approaching. Robert turned his full attention to pulling off an
the very dramatic, evolving scene than a corporate position auspicious event.
would allow. I wanted to see the Web proliferate, not sink my The conference began at C E R N on May 25, and would last
life's hours into worrying over a product release. While leading a three days. It was a tremendous gathering. The auditorium held
consortium would limit my public opinions due to confidentiality perhaps three hundred people. We limited registration to three
and the requirement of having to be neutral, I'd be free to really hundred, but ended up with three hundred fifty after admitting
think about what was best for the world, as opposed to what members of the press, and others who just appearedâ€"testimony
would be best for one commercial interest. I'd also be free to to how the Web had grown.
wield a persuasive influence over the Web's future technical The student volunteers, whom Robert had rounded up to
directions. help run the conference, were manning the registration area.
I suppose I could, as an alternative, have pursued an acade- Robert and I, of course, were running around trying to get the
mic career, gone to a university somewhere as an assistant pro- last-minute things together. But when I went to go into the con-
fessor. But I'd never taken a Ph.D., and so even at C E R N , the ference area, I was very effectively bounced by the students,
grade I had on entry, and the grade I was stuck with throughout because the conference wasn't open yet. It took me a long time to
my career, was one notch down. I would have had to spend a get across to them the fact that I was actually involved with the
good amount of time getting a Ph.D., which would have been in organization that was holding the conference.
a relatively narrow area. I certainly didn't have the time. And As he had promised, Robert had set everything up, and
narrowing my view would have meant jumping off the bobsled I except for the last-minute rushing, I didn't have to do anything
had managed to push into motion. but attend and speak. The environment in the meeting rooms
A more tempting option was to join the research group of a Was exciting yet close. There were people from all walks of life
large benevolent company, which would have allowed me to pur- brought together by their enthusiasm for the Web. Talks given in
sue research that was interesting to me, but also participate in the small auditorium were packed. Because it was the first such

84 85
w e a v i n g t h e w e b c h a n g e s


conference, many people who had been interacting only by e-mail ones now creating the Web, and therefore were the only ones
were meeting each other face-to-face for the hrst time. And for who could be sure that what the systems produced would be
the hrst time people who were developing the Web were brought appropriate to a reasonable and fair society. Despite my trepida-
together with all sorts of people who were using it in all sorts of tion, I was warmly received, and I felt very happy about having
ways. The connections were electric. For example, there was made the point. The conference marked the first time that the

Borre Ludvigsen, who had a home server that allowed people to people who were changing the world with the Web had gotten

visit his house, look at a cutaway model of it, see where the together to set a direction about accountability and responsibility,

computers were in it, and browse his bookshelves. He had put and how we were actually going to use the new medium. It was
an important direction to set at this juncture.
his server on a special phone line provided by the Norwegian
phone company as part of an experiment. He was talking with I went home feeling very pleased. Exciting though all this
people who actually, thought they could adapt his approach for was, in my personal life it was dwarfed by the arrival of our sec-

health-care applications. The excitement, congeniality, and grass- ond child in June. Family life continued and for a while it seemed

roots fervor for furthering the Web inspired the reporters there, MIT had stalled in preparations for the W W W Consortium. Then

overdoing it a little, to dub the meeting the "Woodstock of the Al Vezza began calling me at home in the evening to discuss
details. The conversations seemed even more odd because of the
Web."
cultural disconnect. Our little prefab house was in a small French
In the span of one session in one of the meeting rooms, the
village a few miles from the border with Switzerland. The view
agenda was laid down for H T M L for the next few y e a r s - h o w to
from our front yard stretched straight across Geneva to Mont
incorporate tables, math, and the handling of graphics and photo-
Blanc. From the backyard, where we often ate dinner, was a view
graphic images. Although anything on an Internet F T P server was
of the Jura mountains, cows grazing on the few intervening fields.
available on the Web, H T T P had completely taken off as a more
Given the time difference with Massachusetts, that's often where
efficient alternative, but it needed a lot more optimization to keep
I was when A l called. I would be wearing shorts, sitting out in
up with ever-increasing demands to frequently fetch Web pages
the sunshine. A l , who was certainly wearing a gray suit, would
from a server in rapid succession, and pick up all the graphics
be seated in an air-conditioned concrete office building in Cam-
embedded in a page. In a birds-of-a-feathter session, Dave
bridge. It was sometimes hard to connect across this gulf.
Raggett proposed a "Virtual Reality Markup Language," an idea
Mark Pesce picked up and ran with to start the whole community
One evening in early July our phone rang. It was Al, and he was
doing 3D on the Web and to define V R M L .
serious. He wanted to know if there was a way he could fax me
The only time I felt a bit uneasy was when I gave the closing
nght then and there. He said he had just gotten the go-ahead
speech. I talked about several technical points, which was fine,
rom M I T to form the consortium. L C S was prepared to hire me
announced the upcoming consortium, which was fine. But then
a
« a full-time staff member. He had a letter to that effect, and
finished by pointing out that, like scientists, people in the We^
Wanted to know when I would start.
development community had to be ethically and morally aware
It was just ten days before we were due to leave on our vaca-
of what they were doing. I thought this might be construed as a
»on. We had not specifically planned any dates after that, since
bit out of line by the geek side, but the people present were t
86 87
w e a v i n g t h e w e b
c h a n g e s

the process of getting the details right at M I T seemed at times to
concern that Web technology would move west, leaving Europe
have no end in sight. As it appeared that M I T had now gotten its
behind. I knew I had to move to the center of gravity of the Inter-
ducks in a row, however, there was no reason to wait. September 1
net, which was the United States. The American government
seemed like a .good starting date. It would be only ten days after
could congratulate itself on successful research funding that led
we'd come back from vacation, but we wanted to start in the
to the Internet, and Europe could congratulate itself on taxpayer
States at the beginning of the school year.
money well spent on C E R N .
Al's next call was on July 14, Bastille Day. As usual, our vil-
I left Geneva, off to MIT. Off to America. .Off to the World
lage was celëbrating with fireworks, lit from a field just across
Wide Web Consortium. And off to a new role as facilitator of the
the road from our house. I found that I could not be totally seri-
Web's evolution.
ous with Al, and wondered if he would understand. There we
were, watching the fireworks over our little town in the French
countryside, across the lake from the Alps. The conversation was
almost inaudible with the explosions.
My wife and I were packing our bags for vacation. Although
we assumed we could come back to sort out our affairs, we
decided that if there was a question about whether to bring
something or not, we should bring it. And so we left, with a
young daughter, an infant son, and a cavalcade of friends going
down to the airport with sixteen cases and boxes. My family
never came back. I returned for ten days to sell, with the help of
friends, the cars and the house.
Meanwhile, encouraged by George Metakides in Brussels,
M I T and C E R N inked an agreement to start the World Wide Web
Consortium. It was announced in Boston by Martin Bangemann,
one of the European Commission's commissioners, who was
charged with developing the E C ' s plan for a Global Information
Society. There was a press release. The Associated Press ran a
story about it. Reports followed in the Wall Street Journal, the
Boston Globe, and other major papers. Mike Sendal and Robert
Cailliau had been joined by François Fluckiger, who was to lead
the consortium team at C E R N . It still wasn't clear how the con-
sortium would fit in there, since this was new. It was clear that
M I T was very much in control, moving faster, with more experi-
ence and relevant contacts. Some people in Europe expressed

88
89
 C H A P T E R 8




C o n s o r t i u m




en I arrived at MIT's Laboratory for Computer Science, I
camped out in a corridor with two doors and no windows close
to the offices of Michael Dertouzos and Al Vezza. Though an
office of my own would have been nice, this arrangement actu-
ally worked out beautifully because it allowed us to work
together very readilyâ€"and them to keep an eye on me.
I hadn't had time to get a car yet, so I was commuting by bus
from our temporary home. Trudging to work in citified Cam-
bridge was a far cry from rural France, but it was autumn, and
the bus ride gave me time to revel in New England's fall colors. It
also gave me time to think about my new role.
Although I knew I would be forced to introduce some struc-
ture, I wanted the consortium to operate in a way that reflected a
w
eblike existence. The Web would not be an isolated tool used by
People in their lives, or even a mirror of real life; it would be part
°f the very fabric of the web of life we all help weave.

91
 w e a v i n g t h e . w e b c o n s o r t i u m



- The Web scene was beginning to fill with a colorful mix of events for years to come. In April 1994, Gates had decided that
different types of people, organizations, and concerns. The con- the next version of Microsoft's operating system, Windows 95,
sortium would, too. It would be its own web, and sustain the should include software for accessing the Internet. The decision
greater Web, which would help sustain the web of life. came only a few weeks after Clark and Andreessen formed
I wanted the consortium to run on an open process like the Mosaic Communications. Gates wrote a memo to Microsoft
IETF's, but one that was quicker and more efficient, because we employees saying the Internet would constitute a new and impor-
would have to move fast. I also wanted an atmosphere that tant part of the company's strategy. If Gates h a d made the deci-
would allow individuals, representing their companies or organi- sion two months earlier, would he have hired the same NCSA
:




zations, to voice their personal ideas and find ways to reach com- people that Mosaic had just grabbed?
mon understanding. There would always be people who would The Web was becoming a business. Rather than develop its
disagree, and they would be levers for progress. We would get own Web code, Microsoft licensed browser code from a small
ever closer to true consensus, perhaps never completely achiev- NCSA spin-off called Spyglass. The cost was $2 millionâ€"more
ing it, but delighting in every advance. money than any of us involved from the early days would ever
This freewheeling design might create tension between my have dreamed of.
being a manager and leaving the consortium as a very flat space In November the major marketing campaigns began. At
of peer respect and joint decision-making. It might create tension Comdex, the twice-yearly computer trade show, Microsoft
among consortium members, who would have to take leads on announced with great fanfare that its online service, the Microsoft
issues but always hew to a democratic process. It struck me that Network (or MSN), would be launched and that software to access
these tensions would make the consortium a proving ground for and use it would be part of Windows 95. At the same conference,
the relative merits of weblike and treelike societal structures. I Jim Clark announced publicly that Mosaic Communications was
was eager to start the experiment. changing its name to Netscape. NCSA had been annoyed about
The WWW conferences continued half-yearly at Darmstadt, Clark and Andreessen using its software name, Mosaic, as a prod-
Boston, and Paris, and the academic institutes hosting them uct name, too, and when the two had hired away NCSA's people,
founded the International World Wide Web Conference Commit- NCSA took offense. A n out-of-court settlement was reached, cost-
tee as a nonprofit organization, to continue the series, with Robert ing the upstart company close to $3 million in expenses and other
as president. On the business side, Netscape was working furi- fees, and requiring it to find a new name. Netscape was it.
ously to release the first commercial version of its browser by the
end of the year. Bill Gates and Microsoft, who had shrugged off Al and I were having our own debates over a name for the
the Internet and the Web, were realizing they might be missing a nascent organization, arriving at the World Wide Web Consor-
good party. Gates assigned people to develop a browser. Microsoft tium, or W 3 C for short. Some of the icohs still have a trace of a
was also investigating the development of an online service that "W30" (Organization), which held for a while.
might compete with America Online, CompuServe, and Prodigy While I worked up a technical agenda, A l energetically
The timing of who was developing which technology, and signed up members. The Digital Equipment people who had sur-
who was working with whom, would determine the course of prised me with their visit at C E R N were among the first on Al's
92 93
w e a v i n g t h e w e b c o n s o r t i u m


list of calls. They joined, and people at other companiesâ€"from the Laboratory for Computer Science and sites in Europe and
upstart Netscape to stalwarts like Hewlett-Packard and I B M â€" Asia would produce specifications and sample code, which mem-
quickly followed. bersâ€"and anyone else, for that matterâ€"could pick up and use
Membership was open to any organization: commercial, edu- for any purpose, including commercial products, at no charge.
cational, or governmental, whether for-profit or not-for-profit. Consortium funding from dues (and, initially, public research
The annual fee for full membership was fifty thousand dollars; money) would underwrite these efforts.
for affiliate membership it was five thousand dollars. There was There also would be the Advisory Committee., comprising, one
no difference in benefits, but to qualify for affiliate status an official representative from each member organization, who
organization had to be not-for-profit or governmental, or an inde- would serve as the primary liaison between that organization and
pendent company with revenues less than fifty million dollars. W3C. The committee's role would be to offer advice on the over-
Netscape joined jor the full fifty thousand dollars despite qualify- all progress and direction of the consortium. I would be the con-
ing as an affiliate; it insisted that it join as a big company on sortium's director; A l would be chairman.
principle. Members had to commit to a three-year term of member- Most of the organizations that were signing up were compa-
ship, after which they could renew annually. In return, members nies interested primarily in advancing the technology for their
were free to attend any meeting, and sit on any working group own benefit: The competitive nature of the group would drive the
or other ensemble we would put together. They would also get developments, and always bring everyone to the table for the next
exclusive access to in-depth information on all activities under issue. Yet members also knew that collaboration was the most effi-
way, whether they were directly involved or not. cient way for everyone to grab a share of a rapidly growing pie.
Though we didn't have the motto at the time, the consortium's Although the consortium was seen as primarily an industry
purpose was to "lead the Web to its full potential," primarily by group, the U . S . and European governments were supportive. In
developing common protocols to enhance the interoperability and fact, the U.S. Defense Advanced Research Projects Agency pro-
evolution of the Web. To do this, we would stay ahead of a signifi- vided seed money, in part because we would be building bridges
cant wave of applications, services, and social changes, by fulfilling between academic research and industry. Martin Bangemann, the
a unique combination of roles traditionally ascribed to quite differ- European Commission commissioner, held a meeting of the Euro-
ent organizations. pean governments, which decided to support C E R N ' s coordina-
Like the IETF, W 3 C would develop open technical specifica- tion of Europe's part of the consortium.
tions. Unlike the IETF, W 3 C would have a small full-time staff to Not surprisingly, one of my first steps at M I T was to set up a
help design and develop the code where necessary. Like indus- Web server. I took a copy of all the existing Web documentation
trial consortia, W 3 C would represent the power and authority of and specifications from the info.cern.ch server at C E R N . The
millions of developers, researchers, and users. And like its mem- new web address was http://www.w3.org. C E R N would maintain
ber research institutions, it would leverage the most recent mfo.cern.ch as a forwarding address.
advances in information technology.
The consortium would also take great pains to remain a "ven- N
o sooner had I arrived at MIT than I was off to Edinburgh, Scot-
dor neutral" forum for its members. A small, core staff housed at land, for the next European Conference on Hypermedia Technology.
94 95
w e a v i n g t h e w e b c o n s o r t i u m


It was run by Ian Ritchie of O w l , whom I had tried to convince Meanwhile, Ari Luotonen, the Finnish student from the
four years earlier to develop a Web browser as part of Owl's Erwise project whom Robert had brought to C E R N , was produc-
hypertext product, Guide. It was here that I saw Doug Engelbart tizing C E R N ' s H T T P code. He made it easy to install, with docu-
show the video of his original N L S system. Despite the Web's mentation on how to use it. When his term as a C E R N student
rise, the SGML community was still criticizing H T M L as an infe- came to an end, he joined Netscape to work on its server soft-
rior' subset, and proposing that the Web rapidly adopt all of ware. The other student at C E R N , Henrik Frystyk Nielsen, joined

S G M L . Others felt that H T M L should be disconnected from the us at the consortium. He would be one of the people who would

ungainly S G M L world and kept clean and simple. do the core work on the next upgrade of the hypertext protocol
HTTP 1.1.
Dale Dougherty of O'Reilly Associates, who had gathered the
early Web creators at the first Wizards workshop and other meet- As members signed up for the consortium, they advised us
ings, saw a third alternative. After one session at the conference, about what they wanted to address first. One of the top priorities
a bunch of us adjourned to a local pub. As we were sitting was network security. Information, such as credit-card numbers,
around on stools nursing our beer glasses, Dale started telling sent over the Web needed to be safeguarded. Netscape was par-
everyone that, in essence, the S G M L community was passé and ticularly interested because it had a deal looming with mammoth
that H T M L would end up stronger. He felt we didn't have to MCI to distribute Netscape's browser on MCI's new Internet ser-
accept the S G M L world wholesale, or ignore it. Quietly, with a vice, due to begin in January. Netscape's software, called Secure
smile, Dale began saying, "We can change it." He kept repeating Sockets Layer (SSL), would protect credit-card purchases on
the phrase, like a mantra. "We can change it." MCI's planned online shopping mall. Seeing SSL as a competitive
Right then and there, fixing S G M L was put on the agenda. advantage and feeling that W 3 C was not yet really up and run-
For the H T M L community, the controversy quickly became a ning, Netscape decided not to wait, and developed the software
huge turn-on. It got them going. And many in the documentation fairly independently. This was one of the first programs that
community, also fed up with aspects of S G M L , sympathized. allowed electronic commerce (e-commerce) to gain credibility.

Compared with all the drama taking place in the forming of With so much new, autumn passed quickly. Suddenly it was
Web companies, this controversy may have seemed like an eso- December 1994. In three short days, huge events took place that
teric technical point. But the Jim Clarks and Bill Gateses would would forever alter the Web's future: The consortium members
have no big business decisions to make unless specific decisions met for the first time; Netscape released the commercial version
like the relationship of H T M L to SGML were sorted out. Business- °f its browser; and C E R N decided after all not to be a W 3 C host
people and marketers who thought they were "driving" the Web ^ Slt
e. That bobsled I had been pushing from the starting gate for
would have had nothing to drive. j so long was now cruising downhill.
In October 1994, Netscape released the first version of l t s
|
browser, dubbed Mozilla. It was a "beta" or test version, released &<> | the D C C e m b
Web Consortium held
e r 1 4 a t L C S t h e W o r l d W i d e



people on the Net would try it and send suggestions for improvej ^ first meeting of its Advisory Committee. The meeting was
ments. As he had with Mosaic, Andreessen pumped out message j Co^ f
^ n
a l l , with only about twenty-five people.
e n d l y a n d u i t e s m



about Mozilla over the newsgroups, and users snapped it up. | Petitors in the marketplace, the representatives came together
m




97
w e a v i n g , t h e w e b


with concerns over the potential fragmentation of H T M L . This Whether inspired by free-market desires or humanistic ideals,
was seen as a huge threat to the entire community. There were so vve all felt that control, was the wrong perspective. I made it clear
many proposed extensions for H T M L that a standard really was that I had designed the Web so there should be no centralized
needed. We wrestled over termsâ€"whether the consortium should place where someone would have to "register" a new server, or
actually set a "standard" or stop just short of that by issuing a for- get approval of its contents. Anybody could build a server and
mal "recommendation." We chose the latter to indicate that getting put anything on it. Philosophically, if the Web was to be a univer-
"rough consensus and running code"â€"the Internet maxim for sal resource, it had to be able to grow in an unlimited way. Tech-
agreeing on a workable program and getting it out there to be nically, if there was any centralized point of control, it would
triedâ€"was the level at which we would work. We also had to rapidly become a bottleneck that restricted the Web's growth,
move fast, and didn't want to be dragged down by the sort of long and the Web would never scale up. Its being "out of control" was
international voting process that typihed the setting of an actual very important.
standard. It was becoming clear to me that running the consor- The international telephone system offers a decent analogy.
tium would always be a balancing act, between taking the time to The reason we can plug in a telephone pretty much anywhere in
stay as open as possible and advancing at the speed demanded by the world is because industry agreed on certain standard inter-
the onrush of the technology. faces. The voltages and signals on the wire are almost exactly the
We also decided that if we were going to develop open, com- same everywhere. And given the right adapter, we can plug in a
mon protocols and stay ahead of applications, we would have to wide range of devices from different companies that send all
support an ongoing effort, primarily by the staff, to create a set of sorts of information, from voice to fax to video. The phone sys-
Web tools we could use ourselves to demonstrate new ideas and tem defines what it has to, but then leaves how it is used up to
experiment with proposed specifications. Initially, that meant adopt- the devices. That's what we needed for computers on the Web.
ing a browser and server that were a bit ahead of their time. We On December 15, the day after the first consortium meeting,
agreed to use Dave Raggett's Arena browser and the C E R N server as Netscape released the commercial version of Mozilla, renamed
our test beds. Certainly, we would make these and any other tools Navigator 1.0. It was compatible with Microsoft's Windows oper-
freely available for use by anyone. All people had to do was access ating system,, the X Windows system on Unix, and Macintosh.
the public part of the W3C Web site and download a program. The browser was significant not so much for its technical fea-
Indeed, the true art for the consortium would be in finding tures, but for the way in which Mosaic released it. Rather than
the minimum agreements, or protocols, everybody would need in shrink-wrap and ship it, Netscape released it over the Internet.
order to make the Web work across the Internet. This process did And rather than charge for it, it was free. Within several months
not put the consortium in a position of control; it was just provid- the majority of people on the Web were using it.
ing a place for people to come and reach consensus. In these Andreessen was following the model by which all previous
early days, before we developed more formal processes, if a Web software had been released, except that this time the soft-
member didn't want to be part of a given initiative, the member's ware was coming from a commercial company that was supposed
representative wouldn't come to that meeting. And if people to make money. People wondered where the profit was going to
couldn't agree after serious effort, we'd eventually drop the topic come from.
98 99
 w e a v i n g t h e w e b c o n s o r t i u m


Andreessen and Clark had realized that browsers would board; distribute the software fast and cheap over the Internet;
rapidly become a commodity. NCSA had licensed the Mosaic then try to make money from the millions of visitors through ads
code to other startups, and Microsoft was developing its own r services.
0


browser. Netscape couldn't hope to make its living from the On December 16, 1994, a third day in an incredible week,
browser market. What it could do was get its browser out before CERN announced major news. After negotiating for several years,
the others. If it was rapidly and widely accepted, then the com-
the C E R N Council had unanimously approved the construction
pany would have a platform from which to launch other products
of the Large Hadron Collider, a new accelerator. It would be the
for which it would charge money. It would also bring millions of
next leap toward investigating the even smaller scales of matter. I
people to Netscape's home pageâ€"the default hrst screen when
would soon learn, however, that to accomplish such a mammoth
Navigator was opened. There, Netscape could display ads from
undertaking C E R N would impose stringent budget conditions
companies that would pay to reach a large viewership. The site
across the organization. No program that wasn't central to high-
also would instantly notify browsers of Netscape's other services,
energy physics could be supported. That meant that C E R N ,
which the company would charge for. Netscape also would
regretfully, could not continue to support Web development, or
charge companies for a commercial grade of the browser, which
the consortium.
was more powerful, and for setting up and supporting a com-
In a way, it was probably in everybody's best interests for it
pany's Web server.
to opt out. C E R N , at its heart, had always concentrated on high-,
In taking this position, Netscape was wisely acknowledging
energy physics, and had never developed great experience with
that on the Web, it was more profitable to be a service company
industry or a general policy about working with it. But I felt that
than a software company. Andreessen and Clark may not have
C E R N deserved the credit for letting me develop the Web, and
been completely clear on this at the beginning, though, because
for maintaining such a tremendously creative environment. Con-
people who downloaded the browser were told that they could
tinued involvement in the consortium would have cemented its
use it free for only three months. After that they were expected
place in the Web's ongoing history. I would rather have seen the
to pay, or they would be in violation of the licensing agreement. I
organization get a pat on the back than go quietly into the night.
didn't know what reaction Netscape was getting to this. I
For his part, Robert would remain very involved with the Web
assumed that some people paid, but many did not, and simply
community, by continuing to organize the annual W W W Confer-
downloaded the next version of the software, which also turned
ence series.
out to be free. Netscape allowed this to happen for fear of losing
C E R N ' s resignation left the consortium without a European
fans to other browsers, and as time went on its appeal for pay-
base, but the solution was at hand. I had already visited the Insti-
ment was minimized.
tut National de Recherche en Informatique et en Automatique
This approach set the tone for the Web companies that would
(INRIA), France's National Institute for Research in Computer
follow: Release beta versions for review, which put a nascent
Science and Control, at its site near Versailles. It had world-recog-
software program in the hands of hundreds of professional and
nized expertise in communications: their Grenoble site had devel-
amateur users, who would (for free) send suggestions for
oped the hypertext browser/editor spun off as Grif that I had
improvements; give away basic software to get customers on
been so enamored with. Furthermore, I found that Jean-François
100 101
w e a v i n g t h e w e b


Abramatic and Gilles Kahn, two INRIA directors, understood C H A P T E R 9

perfectly well what I needed. INRIA became cohost of the con-
sortium. Later, in early 1996, we would arrange that Vincent
Quint and Irene Vatton, who had continued to develop Grif
would join the consortium staff. They would further develop the
software, renamed Amaya, replacing Arena as the consortium's
flagship browser/editor. C o m p e t i t i o n
V




The whirlwind of events that had taken place in a mere seventy- a n d C o n s e n s u s
two hours was exciting yet daunting. The consortium had to get
moving with a sense of urgency if it was going to stay ahead of
the large forces that were gathering.
I had to wait only two months for confirmation that the Web
had become a global juggernaut. In February 1995 the annual
meeting of the G7, the world's seven wealthiest nations, was held
in Brussels. The world's governments were rapidly becoming
aware of the technology's influence, and Michael Dertouzos
LCS's director, was invited to join the U.S. delegation there. As
Michael describes in his book What Will Be, the keynote speaker
was Thabo Mbeki, deputy president of South Africa. Mbeki
delivered a profound speech on how people should seize the new
History often takes dramatic turns on events that, at the time,
technology to empower themselves; to keep themselves informed
seem ordinary. Microsoft wanted to license Netscape's browser,
about the truth of their own economic, political, and cultural cir-
buy a share of the company, and take a seat on Netscape's board.
cumstances; and to give themselves a voice that all the world
In return, Netscape would be the browser on Microsoft's Win-
could hear. I could not have written a better mission statement
dows 95, an entirely new operating system, which would launch
for the World Wide Web.
Netscape into the huge personal computer industry. But Jim
Clark and Netscape's new C E O , Jim Barksdale, who had been
hired to raise money and make deals, were wary. The proposal
fell through, and Microsoft redoubled its efforts to offer its own
browser.
Other deals, however, did go through, further shaping the
eompetitive landscape. In April, Compaq announced that its new
hne of personal computers would come with Navigatorâ€"the first
time a browser would be bundled directly with hardware.
102
103
 w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s



In May, with little fanfare, Sun Microsystems introduced o W e r of the large software companies, like Microsoft, since pop-
Java, a new programming language. Java was a repackaging o f : ular software programs such as word processors could be gotten
James Gosling's Oak language, originally designed for appli . \ Ca
in J a v a r a t n e r t h a n f r o m t h e
shrink-wrap market. Java also meant
tions such as phones, toasters, and wristwatches. Small applica- that people with all sorts of different pocket devices, which
tion programs written in Java, called applets, could be sent couldn't support a lot of hardware or software, could communi-
directly between computers over the Internet, and run directly cate and work with each other over the Web from anywhere.
inside a Web page on a browser. That was the theory. It met the
need for applications in which a hypertext page was not suffi- Meanwhile, great anxiety was growing among a^ group of technol-
ciently interactive, and some programming on the client was nec- ogy companies that for several years had been leading the way
essary. The excitement was that even if computer A and toward the Information Age: the online service providers. Compu-
computer B had different operating systems, an applet written on Serve, Prodigy, America Online, and others that offered prepack-
computer A could run on computer B, because the Java language aged content such as news, an encyclopedia, travel information,
set up a virtual computer on computer B that required only mini- and e-mail tended to represent the Internet as some "other" net-
mal support from computer B's operating system. Many lan- work that was arcane and complex, certainly not worth hassling
guages, however, had tried to achieve this goal in the past, but with. But the Web suddenly made the Internet easy. It also
the effort of standardizing all the facilities they needed was often enlightened subscribers to the fact that these online companies
their demise. were either isolated islands or just a small part of the Internet. To

Initially, Java worked. Suddenly, a professional or amateur keep their customers, the online service providers grudgingly

programmer could create a Java application, post it on a Web site, provided access to the Web, though they still tried to represent it

and people everywhere could download and use it. Java opened as something that was part of their kingdom. As press coverage

up a wide world of potential Web applications that would be of the Web increased, the services became more careful about

simple and inexpensive. Netscape immediately licensed Java, and not misrepresenting the Web to a smarter public. They had to

incorporated it into its next version of Navigator. I was very reverse their stance, repositioning themselves as providing orga-

excited because Java is an object-oriented language, a more pow- nized and safe content, so that people didn't have to venture out

erful programming technique that I had used to write "World alone onto the Web to find what they wanted.

Wide Web" but had abandoned due to lack of standardization. As part of the general upheaval, America Online (AOL)
In theory, a computer would not need a substantial hard disk bought Navisoft, the company that had developed the Navipress
and working memory (RAM) to store and run volumes of soft- browser that also worked as an editor. A O L changed the product
ware for various applications such as word processing, bookkeep- name to AOLpress. (It is the software that I used to draft early
ing, and the like. Instead, a computer with minimum memory parts of this book.)
and R A M could call up a Web site and download a Java applet for At one point, there were even rumors that A O L was trying to
writing documents or keeping books. Personal computers could start a consortium like W3C, with a similar name. I sent an e-mail
therefore be made with less hardware and thus at a lower price. to AOL's chief executive, Steve Case, to try to bridge the cultural
Some people even thought this new development could erode the gap. They gave up on the idea, realizing that all the Web companies
104 105
w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s


were already part of W3C, and were far too big a group for them After Netscape's IPO, people began to ask me whether I was
to try to control. upset by the Web "going commercial." They still ask today. One
Realizing that Netscape had to grow fast if it was going to part of the question means: "Are you upset that people have to
compete with the big guys like Microsoft, Netscape's chief execu- pay money for certain Web products, or at least for commercial

tive, Jim Barksdale, decided the company should go public, to get support for them?" Of course I am not. The free software com-
munity was fundamental to the development of the Web, and is a
a big cash infusion. The initial public offering (IPO) was held on
source of great creativity. But it was inevitable and important
August 9, only sixteen months after the company was formed.
that if the Web succeeded, there would be a variety of free and
This was extremely early for an IPO, but Wall Street was paying
commercial software available.
premium prices for high-technology stocks, and Netscape needed
ammunition to compete with Windows 95 and the browser that A second meaning to the question related to the fact that for
would come with it, which were due out very soon with heavy a long time Web pages were posted by individuals and not-for-
Microsoft promotion. profit organizations, which pointed to each other with no thought

The stock was set to open at twenty-eight dollars a share, of commercial gain. Academics who had used the Internet from
its early stages felt it was an open, free, pure space for their use,
already a high price, but demand rapidly pushed it to seventy-
and they worried that the bountiful information space they had
one dollars. Morgan Stanley, the investment house managing the
enjoyed for these righteous uses would now become unavailable,
offering, could not issue shares fast enough. Scores of large insti-
swamped by junk mail and advertising. Certain people felt that
tutions wanted large percentages of ownership. They kept buying
commercially motivated material polluted the Web. I had little
more until, at the close of trading, 38 million shares were on the
time for this point of view. The Web was designed as a universal
market. Netscape, after a single day of trading, was worth $4.4
medium. A hypertext link must be able to point to anything.
billion. It was the largest IPO in history, and the company had
Information that is put up for commercial gain can't be excluded.
yet to show a profit.
People have sometimes asked me whether I am upset that I
If the World Wide Web had not yet gotten the public's full
have not made a lot of money from the Web. In fact, I made
attention, this remarkable story put it on center stage. It also sent
some quite conscious decisions about which way to take my life.
an undeniable message to the commercial world: The Web was
These I would not change-though I am making no comment on
big business. The gold rush was on. The flood of cash enabled
what I might do in the future. What does distress me, though, is
Netscape to buy small companies that had developed specialized
how important a question it seems to be to some. This happens
products for the Web, create joint ventures with larger corpora-
mostly in America, not Europe. What is maddening is the terrible
tions, and broaden its product line to support big contracts from
notion that a person's value depends on how important and
major corporate buyers. By the end of 1996, when it settled into financially successful they are, and that that is measured in terms
its full business model, Netscape would employ more than two 'j of money. That suggests disrespect for the researchers across the
thousand people and report revenues of $346 million. Its inflated % globe developing ideas for the next leaps in science and technol-
stock price would come down to reasonable levels over the com- ogy. Core in my upbringing was a value system that put mone-
ing years, but in one fell swoop the Web had become a major tary gain well in its place, behind things like doing what I really
market. a
. 1 0 7
106
w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s

want to do. To use net worth as a criterion by which to judge ing about it constantly. Internet service providers, ISPs, sprouted
people is to set our children's sights on cash rather than on things everywhere, offering Web access for twenty to twenty-five dol-
that will actually make them happy. lars a month. Computer jocks in small towns around the globe
It can be occasionally frustrating to think about the things my started putting up their own homepages, and soon enough
family could have done with a lot of money. But in general I'm offered to do the same for businesses, mom-and-pop stores,, and
fairly happy to let other people be in the Royal Family role (as it individuals.
were), as long as they don't abuse the power they have as a The consortium had positioned itself to help the Web move
result. The consortium is the forum where people setting the positively forward. We were holding meetings and issuing brief-
agenda meet. It's not as if I can just make decisions that change ings packages. But our head of communications, Sally Khudairi,
the Web . . . but I can try to get an entire industry organization to realized we needed more than an efficient Web site to get our
do it. My priority is to see the Web develop in a way that will message across. She rapidly set up relationships with the press
. hold us in good stead for a long time. If someone tries to monop- and channels to all those we needed to tell about W 3 C work. The
olize the Webâ€"by, for example, pushing a proprietary variation members suddenly found out all kinds of things about their con-
of network protocolsâ€"they're in for a fight. sortium they never knew, and people who really needed to know
about W 3 C Recommendations but had never heard of us were
Two weeks after Netscape's IPO, Microsoft released Windows 95, soon using our name as a household word.
and with it Microsoft's browser, Internet Explorer. Bill Gates was Al Vezza was an effective chair and in essence C E O for the
turning his back on his earlier strategy of creating a dial-up ser- first years; he was succeeded by INRIA's Jean-François Abra-
vice, the Microsoft Network, patterned after AOL. matic, whom I had met when I first visited INRIA. Alan Kotok,
The first version of Internet Explorer had very little function- who was one of the four people from Digital Equipment who had
ality. I could tell it was put together in a hurry, but it got shown up at my office in Geneva, ended up being on the Advi-
Microsoft's toe in the water. In December 1995, Gates made what sory Committee, and is now on the staff as associate chair. Dale
would later be seen as a famous speech to the press, in which he Dougherty, who chanted, "We can change it" in that Edinburgh
announced that his company was going to "embrace and extend" bar, would later join the Advisory Board, a small group elected
the Internet. To certain people in the computer industry, from the full Advisory Committee.
"embrace" meant that Microsoft's products would start off being The consortium soon began to develop and in turn codify its
compatible with the rest of Web software, and "extend" meant Process for developing future technology and recommendations.
that sooner or later, once they had market share, Microsoft's prod- From then on the process would continously evolve and be
ucts would add features to make other people's systems seem refined. Any member could raise the idea of pursuing an issue.
incompatible. Gates was turning the company around very rapidly Members or staff would draw up a briefing package, which
and forcefully, to fully exploit the Web. The business community explained why it was important to address a certain matter. It
was impressed that Gates was getting into this so personally. Would address what the market conditions were, the technical
By mid-1996, millions of people were accessing the Web, ls
sues, why the consortium rather than someone else should
thousands of companies were serving it, and the press was writ- ta
ckle this, how • we could help the situation, what the next step
108 109
w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s



would beâ€"a workshop, a working group, a slew of working One day Dan Connolly arrived very disgruntled at the con-
groupsâ€"and how much it would cost us to pursue. sortium staff's regular Tuesday meeting at L C S . I had met Dan
A briefing package would be distributed to the whole mem- w a y back at the hypertext conference in San Antonio where
bership. Members would review the package, returning com- Robert and I had soldered together the modem so we could
ments as to their support and likely participation. If there was demonstrate the Web. A red-haired navy-cut Texan, Dan had
sufficient support and no serious problems, we would most often been very active on the Internet and was an expert in many areas
create a new activity. Activities could contain any number of key to Web technology, including hypertext systems, and markup
working groups, coordination groups, interest groups, and staff so languages. He had since joined the W 3 C staff and was leading
as to get the job done in an open, high-quality, and efficient way. our Architecture domain. O n this day, he came in saying the
In addition to considering the core technical issue, the con- consensus process had broken down in a working group, and all
sortium had to consider the impact on the society being built hope of meeting the deadlines promised to other groups seemed
over the Web, and political questions such as whether govern- lost. One company was becoming a big problem, though • he
ments were likely to do rash things if a technology was not devel- couldn't tell for exactly which reasons. The specification wouldn't
oped correctly. With every new activity, the mix of pressures be able to come out, and the failure would be a blow for the con-
would be different. The consortium had to be able to respond in sortium and the Web community.
a very flexible way to put together a structure and strategy that Dan didn't really want to talk about it, but the rest of the
were appropriate. team dragged him back to the subject. This sort of problem was
Working groups could offer their specifications for wider and the crux of the job. Technical issues might be more fun, but this
wider review by other groups, the membership, and the public. was the stuff of building consensus, of making progress in an
The final phase occurred when a solution became a Proposed Rec- open community.
ommendation, up for formal member review. All the members Did the problem company really not want to agree? Was
then would be asked to comment within thirty days. It would there no way to arrive at consensus? Each of us interrogated Dan.
either become a W 3 C Recommendation, be sent back for changes, We diagrammed what was happening on the whiteboard. The
or be dropped altogether. In theory, the outcome was my decision, whole staff worked through it with him. By the end of the meet-
based on the feedback (much as the monarch, in theory, rules ing, Dan and the team had developed a way to bring the spec for-
Britain!), but in fact we would put the member review comments ward. The companies agreed within two weeks. It was rewarding
through an internal process of review with the domain and activ- for me to see that the process worked even in times of contro-
ity leads and working-group chair. In most cases there would be versy, and it meant a great deal to me that the staff could work so
clear consensus from the membership anyway. In a few cases we well together.
would go ahead despite objections of a minority, but then only Of course, at times there was tension when people from dif-
after having delivered a detailed analysis of the opinion overruled. ferent companies had different technical views on how to settle a
Once a Recommendation was passed, the membership was recommendation. It was often difficult to predict which company
informed, a press release would go out, and Sally's PR machine • representative might play the good guy or bad guy. But finding a
would encourage everyone everywhere to adopt it. technically sound, common solution was the job we were about.

110
Ill
w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s


Indeed, the consortium thrived on the tensions. The competitive their time browsing the Web, and a large proportion of what they
struggles for chunks of a lucrative market now provided the w e r e viewing was pornography.
financial backdrop for the technological revolution, which itself Exaggerated though this take on the situation may have been,
was the backdrop for a real social revolution. Everyone had a a group of companies quickly came to the consortium asking to
common need to see that the technology evolved. do something now, because they knew Congress had plans to
draw up legislation very soon that would be harmful to the Inter-
During 1996, Netscape released Navigator 2.0, which had easy- net. Already, Web sites acceptable to people in Finland were
to-use e-mail and supported Java applications. Bit by bit, the appalling to people in Tennessee, and the idea of Washington try-
online service providers were giving up and providing gateways ing to decide what was "indecent" for everyone in the world was
to the Web. Bill Gates agreed with AOL's Steve Case to provide indeed sinister.
AOL with a version of the Explorer browser so that A O L sub- The consortium companies realized that as an industry they
scribers who accessed the Web through AOL's gateway could had to. demonstrate that they could produce a solution. They had
browse. An unfortunate outcome of this arrangement, however, to show that, with simple technology, they could give parents the
was the death of AOLpress, one of the few commercial browsers means to control what their children were seeing, with each par-
that provided simple online editing. ent using their own definition of what material was appropriate,
The consortium's biggest social test came in response to pos- not Washington's. The idea was to create a simple program that
sible government overreaction to the public's rapidly rising con- could be installed on or in any browser and would let parents
cern about pornography on the Web. John Patrick from I B M was block the display of sites that carried a certain rating, like the "R"
the first W 3 C member to broach the topic. Sitting to one side of or "X" rating of a movie. However, the program would allow par-
the small room at L C S at that first meeting of twenty-five people, ents to choose from any number of rating schemes that would be
John mentioned that there might be a problem with kids seeing devised by different commercial, civic, even governmental groups.
indecent material on the Web. Everyone in the room turned A rating service would simply be found at the group's U R L
toward him with raised eyebrows: "John, the Web is open. This is The consortium would define the languages for writing the
free speech. What do you want us to do, censor it?" ratings and for serving them up on the Web. We called this work
Underlying his concern was the fact that I B M was trying to the Platform for Internet Content Selection (PICS) and released it
install computers in classrooms. across America, and it was to the public in March 1996. Member companies would incorpo-
meeting with resistance because parents and teachers were wor- rate the technology into their products.
ried about access to inappropriate material. "Something has to The legislation everyone was terrified of surfaced as the Com-
be done," he maintained, "or children won't be given access to munications Decency Act, which rode on the big Telecommunica-
the Web." tions Act that was certain to be passed. Proposed by both the
This was a sobering and new concern for many of us. We Democratic and Republican parties, it would regulate content on
decided to return to the topic at a later meeting, but then Time the Net. We rapidly promoted PICS, and a number of the compa-
magazine published Marty Rimm's article alleging more or less nies that had members on the PICS working group funded press
that a large proportion of students spent a large proportion of events. The Communications Decency Act passed, but then civil
112 H3
 w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s



rights groups challenged it in the courts. Ultimately, it was over- I went down to the basement of a local television studio,
thrown as unconstitutional. The existence of PICS was an impor- where I was going to be hooked up so I'd appear to viewers as a
tant factor in helping the courts see that the act was inappropriate guest in a window on the television screen. There I sat, in this
that protection could be provided without regulation and in a
g y windowless box of a room, waiting for the slot to come on
ra



manner more in keeping with the Bill of Rights. t he air. There was an unmanned camera pointing at me, and a
Ratings schemes were subsequently devised, and a number of television monitor that showed the program in progress. My ris-
companies incorporated the technology. Other companies that ing unease with the situation suddenly spiked when*I heard the
specialized in child-protection software sprang up. But the furor anchor break in and say, "We'll be back in a few minutes with
calmed down, people relaxed, and industry didn't push PICS Tim Berners-Lee, and his plans to control the Internet."
technology. Still, PICS had shown that the consortium could From there it only got worse. When the anchor came back to
work very rapidly, effectively, and in a new arenaâ€"the overlap- start the segment with me the monitor went blank. I tried to con-
ping area of technology, society, and politics. centrate on the anchor's voice in my ear and the camera in front
of me, with no visual clues as to what was going on. Suddenly,
Just after the consortium released PICS, I made the mistake of they cut me in. The anchor's first words were: "Well, Tim Berners-
talking about it to a reporter who found the principle difficult to Lee, so you actually invented the World Wide Web. Tell us, exactly
understand. I thought it was rather simple: W 3 C develops the how rich are you?"
protocols, some other party develops the rating schemes, other Clearly, the fine points of PICS were not what they were
parties like civic groups would issue ratings, the protocols would after. I was flummoxed. They were annoyed, then eager to hustle
be incorporated into commercial products, and parents would me off as the milliseconds fled by. My debut as a talking head
choose which rating scheme and levels they would use to block was a disaster. Since then, I have not been eager to return to live
material for each child. Combining this with the conditions on television. The next day, as the botched news-wire article made
W3C's sample code, the reporter translated it into the statement ever-wider rounds, there was a large outcry from software com-
that W3C was producing a product for safe Web surfing that panies that we were undercutting their market by (supposedly)
would be distributed free to all parents, and by the end of the releasing competitive products for free. We fought a hard rear-
year! The story suggested that W3C would be undermining the guard action to explain how the story was totally wrong. But this
market for child-protection software. Although it ran in a small, was a big headache we didn't need. I had learned how difficult it
local paper, that paper belonged to a syndicated news wire, and, is to determine what a reporter does and does hot understand,
unbeknownst to me, the story showed up all over the place, even and how vital it is to get one's story across in no uncertain terms.
internationally. I had also learned the fundamental truth about life at W 3 C : We
The next afternoon, still unaware of the article, I got a phone never would know when it would be a quiet day or when the
call from Market Wrap, a fast-paced daily financial program on phone would be ringing off the hook.
C N B C . They asked me if I would answer a few questions for the
evening's program. Acting on the mistaken believe that all public- More companies from Japan and the Pacific Rim were joining the
ity is good publicity, I agreed, consortium, enough so that there was a need for an Asian host.
iiz, 115
 c c m c e t i t i o n a n d c o n s e n s u s
w e a v i n g t h e w e b


Keio University i n japan filled the bill, becoming our third host earlier on potential antitrust violations. It had more recently
issued a consent decree that forbade tight product integration.
institution, w i t h Professor Nobuo Saito as associate chair and
Was Explorer 4.0 truly integrated, or just another bundle?
Tatsuya Hagino as associate director for Japan. Suddenly, finding
U.S. Attorney General Janet Reno announced that the Justice
a good time for global telephone conferences became even more
Department w o u l d take Microsoft to court, on charges of violat-
difficult.
ing the decree. Investigations, injunctions, and hearings w o u l d
The Web industry was growing. The browser companies such
extend the case into 1999.
as Netscape were broadening into server software, and Web
intranets for corporations. Hundreds of large companies, from Whatever the merits of the Department of Justice case, inte-

Chrysler to Federal Express, were starting Web operations. Con- grating a browser w i t h an operating system was connected w i t h

ventional groupware products, such as Lotus Notes, w h i c h had the consistency of user interface for local and remote informa-

been taken over by I B M , were reconfigured so they could be tion. Back at the Boston Web conference i n December 1995, I

accessed w i t h a browser and used to create a Web site. had argued that it was ridiculous for a person to have two sepa-
rate interfaces, one for local information (the desktop for their
Through the consortium's work, HTML steadily became
own computer) and one for remote information (a browser to
more robust. We built on various early work, such as Dave
reach other computers). W h y did we need an entire desktop for
Raggett's handling of tables and figures i n his Arena browser,
our o w n computer but get only a w i n d o w through w h i c h to view
Marc Andreessen's handling of images embedded i n the text of
the entire rest of the planet? Why, for that matter, should we
Mosaic, and style sheets for different fonts and formatting that
have folders on our desktop but not on the Web? The Web was
Hákon Lie had championed since the early days and taken far
supposed to be the universe of all accessible information, w h i c h
beyond the crude f o r m i n m y original browser on the NeXT, as
included, especially, information that happened to be stored
w e l l as new innovations. By mid-1997 Web sites routinely carried
locally. I argued that the entire topic of where information was
beautiful photographs, animated graphics, tabular information,
physically stored should be made invisible to the user. This d i d
audio, and order forms. Hypertext glued them all together i n a
not, though, have to imply that the operating system and browser
multimedia sensation. Though less visible, development of better
should be the same program.
servers was advancing just as quickly.
By autumn, Microsoft's Internet Explorer had garnered a third The Justice Department wasn't concerned w i t h the merits of
of the browser market. But the company turned heads when it software design. The question it raised was whether or not

began to promote its new operating system, Windows 98, sched- Microsoft was using its market dominance to destroy competi-

uled for release i n the spring of 1998. According to Microsoft, this tion. By including the browser w i t h Windows 98, it maintained,
the company effectively eliminated any reason for anyone to pur-
new version w o u l d include an upgraded browser, Explorer 4.0.
chase Netscape Navigator.
The browser would no longer be a program that came bundled
w i t h the system's software, b u i would be an integrated part of the In January 1998 Netscape made a surprise move reminis-
operating system, one and the same w i t h the program that ran the cent of the original Internet ethos: Rather than just giving away
Windows desktop. T h w piqued .he interest of the U.S. Depart- the compiled code for its browser, it said it w o u l d make all the
ment of Justice. The DOJ had investigated Microsoft a few years source code â€"the original text of the programs as w r i t t e n by the

ii6 H7
w e a v i n g t h e w e b c o m p e t i t i o n a n d c o n s e n s u s


programmersâ€"completely public. This open source policy meant a nd compliance, then check its newest product to see if the com-
that anyone promoting a new technology could create their own pany is delivering on those promises. Vendors are driven by buy¬
version of Navigator for it. It meant that any student doing e r s and buyers are largely driven by the press, which can lay
research or simply a class project could create his or her own into anybody it feels is playing a game. The consortium, the
versions of specific parts of the browser, and regenerate Naviga- press, and the user community all work as part of a cycle that
tor with his or her own ideas built in. It meant that anyone who helps the public make reasonable judgments about how honest a
was infuriated by a Navigator bug that Netscape didn't fix could company is being with them.
fix it themselves, and send the fix to Netscape if they wanted, for One of the major technical advances to come from the consor-
future versions. The open release would allow thousands of people tium is a simpler language to supersede SGML, called XMLâ€"the
to improve Netscape's products. Microsoft was bigger than Net- Extensible Markup' Language. Like SGML, X M L is a base for
scape, but Netscape was hoping the Web community was bigger defining languages like H T M L . Dan Connolly, a Web architect
than Microsoft. from early days, had an understanding of the SGML tradition. Jon
Bosak came from a tradition of SGML in ISO committees but saw
The Netscape and Microsoft stories made for dramatic reading, that the Web needed something cleaner. They formed the nucleus
so they were the constant focus of the press. But they were only of what had seemed such a remote hope when Dale Dougherty
a small part of the Web story. By its nature, the work at the con- had muttered, "We can change it," in that Edinburgh pub.
sortium took a much lower profile, but it stuck to the evolving The X M L revolution that followed has been greeted with great
technology. The Web is built on technical specifications and enthusiasm, even by the SGML community, since it keeps the
smooth software coordination among computers, and no market- principles of SGML in place. When Tim Bray, editor of the X M L
ing battle is going to advance either cause. specification, waved it at the attendees at the WWW6 conference
By the end of 1998 the consortium had produced a dozen in April 1997, he was greeted with applauseâ€"because the spec
Recommendations. W3C's technical strength was broader. There was thin enough to wave. X M L has gone on to become one of the
were more than three hundred commercial and academic mem- most widely known of W3C's activities, and has spawned books,
bers worldwide, including hardware and software vendors, conferences, and a nascent X M L software industry.
telecommunications companies, content providers, corporate The consortium has also developed its own set of advanced
users, and government and academic entities. Advisory Commit- Web tools, which we use to test proposed technology as it is
tee meetings had moved from meeting rooms to a large audito- brought to the group. It tries to use its limited resources to
rium, with questions coming from attendees standing at develop at the leading edge where others have not yet ventured.
microphones posted in the aisles. We can't do this all the time, but we have some pretty good
The consortium has learned how to let the outside world put minds at work, and good links with all the major companies and
pressure on a member that may not be acting in an open manner. universities.
We produce Recommendationsâ€"not Standards or regulationsâ€" In 1996 we negotiated the right to the Grif code from INRIA
and we have no way to require anybody to abide by them. But and renamed it "Amaya." It is designed completely around the
journalists can look at a company's statements about openness idea of interactively editing and browsing hypertext, rather than
118 • 119
w e a v i n g t h e w e b
c o m p e t i t i o n a n d c o n s e n s u s

simply processing raw incoming H T M L so it can be displayed on behind the scenes. Jigsaw has had great success as a development
the user's screen. Amaya can display a document, show a map of and test platform among the Java and H T T P cognoscenti, because
its structure, allow the viewer to edit it, and save it straight back the server is so flexible.
to the Web server it came from. It is a great tool for developing Written into the consortium's constitution is the stipulation
new features, and for showing how features from various text- that all the software it produces in support of its work be avail-
editing programs can be combined into one superior browser/editor, able to the public. This is a way of promoting recommendations,
which will help people work together. I switched from AOLpress discussion, and experimentation. It allows anyone to join in the
to Amaya. testing of new protocols, and allows new companies to rapidly
One Web server we use is Apache. When NCSA was develop- get into the swing of Web software creation. All anyone has to do
ing Mosaic, they called me at one point and asked if I would is go to the consortium's site, www.w3.org, and download these
mind if they made a server. My policy, of course, was that I tools for themselves.
wanted as many people as possible writing Web software, so I The consortium's world does sometimes fill up with p o l i t i c s -
said, "Of course, go right ahead." What they meant, but left industrial and governmental. Companies occasionally make tech-
unsaid, was that they'd be writing another server that would be nical statements for commercial reasons. Marketers tamper with
competing for "market share" with the server I had written. But the facts and confuse the public as they fence with the others in
NCSA's subsequent development slowed down, so a bunch of the field. But underneath, the consortium's members are still
people from all over the Net got together to create "patches" for pursuing exciting technological advances. Engineers move from
NCSA's server, and the result, Apache, became a server in its company to company, sometimes with projects their employers
own right. It was maintained by a distributed group of people on are abandoning due to lack of understanding, sometimes leaving
the frontier of Web development, very much in the Internet style. a trail of claims to their ideas made by each place where they
Apache to this day has a huge number of users, and is a powerful worked. The web of life continues to grow in all this activity. And
and flexible server systemâ€"again, a tremendous testimony to the despite commercial pressures, the technical ideas, the consor-
whole idea of open-source software. tium's principles, and the social motivations behind them con-
We use Apache as our main server that is accessible to the tinue to hold center stage.
public. We use our open source "Jigsaw" server for collaborative
editing of all kinds of documents, from W 3 C Recommendations
to our meeting minutes. Jigsaw is a Java-based server, originally
written for the consortium by Anselm Baird-Smith, a slight,
enthusiastic French wizard who can write code at lightning
speed. Anselm wrote Jigsaw initially as background exercise to
help him get used to Java and HTTP. In the two months before he
actually joined the consortium staff he had already rewritten i*
four times. Jigsaw allows members and staff to read and write
documents back and forth, and to keep track of all changes

120 121
 C H A P T E R 10




W e b oF P e o p l e




The Web is more a social creation than a technical one. I
designed it for a social effectâ€"to help people work togetherâ€"and
not as a technical toy. The ultimate goal of the Web is to support
and improve our weblike existence in the world. We clump into
families, associations, and companies. We develop trust across
the miles and distrust around the corner. What we believe,
endorse, agree with, and depend on is representable and, increas-
ingly, represented on the Web. We all have to ensure that the
society we build with the Web is of the sort we intend.
When technology evolves quickly, society can hnd itself left
behind, trying to catch up on ethical, legal, and social implica-
tions. This has certainly been the case for the World Wide Web.
Laws constrain how individuals interact, in the hope of allow-
m
g society to function. Protocols define how computers interact.
These two tools are different. If we use them correctly, lawyers do
n
° t tell computer programmers how to program, and programmers

123
 w e a v i n g t h e w e b w e b of p e o p l e



do not tell legislators how to write laws. That is on an easy day o While these maneuvers certainly affect the business of the
a difficult day, technology and policy become connected. The We y/eb in the larger picture they are the background, not the theme.
Consortium tries to define protocols in ways that do not constrain Some companies will rise, some will fall, and new ones might
the norms or laws that govern the interaction of people. We d e f i ne
s ring from the shadows and surprise them all. Company fortunes
mechanism, not policy. That said, it is essential that policy and and organizational triumphs do not matter to our future as Web
technology be designed with a good understanding of the i m p l i .
ca
users nearly as much as fundamental sociotechnical issues that
tions of each other. As I noted in closing the first International^ could make or break the Web. These have to do with information
C
World Wide Web Conference at C E R N in May 1994, technologists quality, bias, endorsement, privacy, and trustâ€"fundamental val-
cannot simply leave the social and ethical questions to other ues in society, much misunderstood on the Web, and alas highly
people, because the technology directly affects these matters. susceptible to exploitation by those who can find a way

Since the Web is a work in progress, the consortium seeks to- 5 Bias on the Web can be insidious and far-reaching. It can
have a dialogue with policy makers and users about what sort of ' break the independence that exists among our suppliers, of hard-
social interactions the Web should enable. Our goal is to assure ware, software, opinion, and information, corrupting our society.
that the Web accommodates the maximum diversity of public We might be able to hold bias in check if we all could judge the
policy choices. In areas like freedom of expression, privacy, child content of Web sites by some objective definitions. But the
protection, intellectual property, and others, governments do process of asserting quality is subjective, and is a fundamental
have a role. The kinds of tools we make available can help assure right upon which many more things hang. It is asserted using
that those laws are effective, while also ensuring that individuals systems of endorsement, such as the PICS protocol the consor-
retain basic control over their online experience. tium developed to show that government censorship was not nec-
Through 1996, most of what happened to the Web was driven essary. The large number of filtering software tools now available

by pure excitement. But by 1998, the Web began to be seen as a show that government censorship is not even as effective: A

battleground for big business and big government interests. Reli- nation's laws can restrict content only in that country; filters can

gious and parental groups began to call for the blocking of offen- block content no matter where it comes from on the Web. Most

sive material on the Web, while civil rights groups began to important, filters block content for users who object to it without

object strongly to these objections. For this reason, among others, removing the material from the Web. It remains available to

many people in business, government, and society at large would those who want to see it.

like to "control" the Web in some way. I would like to see similar endorsement techniques used to
Unfortunately, these power plays are almost all we hear about express other subjective notions such as academic quality.
in the media: the Justice Department's antitrust case against The essence of working together in a weblike way is that we
Microsoft, the merger mania and soaring stock prices of Internet function in groupsâ€"groups of two, twenty, and twenty million.
companies, and the so-called battle of the portalsâ€"the attempts We have to learn how to do this on the Web. Key to any group's
by mammoth Web sites such as Yahoo!, service providers like existence is the integrity of the group itself, which entails privacy
America Online, and content companies like Disney to provide and confidentiality. Privacy involves the ability of each person to
the widest window to the Web's content. dictate what can and cannot be done with their own personal

124
w e a v i n g t h e w e b w e b o f p e o p l e


information. There is no excuse for privacy policies not to be con- operator error on one of them did once black out the system,
sensual, because the writing, checking, and acceptance of such causing huge disruption. That technical weakness is itself less of
policies can all be done automatically. a concern than the social centralization that parallels it.
Agreements on privacy are part of the greatest prerequisite. Both the domain names and the Internet addresses are given
for a weblike society: trust. We need to be able to trust thé mem- out in a delegated way. To set up the name www.lcs.mit.edu, one
bership of groups, the parties engaging in e-commerce, the estab- registers it with the Lab for Computer Science, which is owner of
lishment of who owns what information, and much more. the lcs.mit.org domain. L C S got its domain name in turn from
Nowhere is the difference between the old tree-oriented model MIT, which is the registered owner of mit.edu. M I T got its
of computing and the web model more apparentâ€"and nowhere domain from the owner of edu. Control over the "top-level"
is society so completely tied to technologyâ€"as the online struc- domains such as,.com and .edu indirectly gives control over all
ture that decides who and what we trust. The criteria a person domain names, and so is something of great power. Who should
uses to assign trust can range from some belief held by their exercise that power?
mother to a statement made by one company about another. During the entire growth of the Internet, the root of an Inter-
Freedom to choose one's own trust criteria is as important a net address was administered by a body known as the Internet
right as any. Assigned Numbers Authority. IANA was set up, was run by, and
A key technology for implementing trust is public key cryptog- basically was the late Jon Postel, an Internet pioneer and guru at
raphy (PKC), a scheme for encoding information so no one else the University of Southern California. Jon managed IANA as a
can read it unless he or she has the key to decode it. How we can public trust, a neutral party. Much of the growth of the Web and
use it directly affects what we can do socially. With this tool, we Internet depended on his integrity as the ultimate trusted author-
can have completely confidential conversations at a distanceâ€" ity who saw to it that the delegation of domain names was fair,
vouch for the authenticity of messages, check their integrity, and impartial, and as unfettered as possible. Because of the sort of
hold their authors accountable. However, it is not available, person Jon was, it worked. The Web and Internet as a whole owe
largely for political reasons explained in the next chapter. a lot to Jon, who died in October 1998 at age fifty-five.
Potential problems of unfair control over domain names
For all its decentralized growth, the Web currently has one cen- loomed larger when the U.S. government decided in late 1998
tralized Achilles' heel by which it can all be brought down or that IANA should be privatized. The potential problem was exac-
controlled. When the U R I such as http://www.lcs.mit.edu/foo is erbated by U R I prospectors. The registration of domain names
used to find a web page, the client checks the prefix, and when, had always been done on a first-come, first-served basis. Increas-
as often, it is "http" it then knows that the www.lcs.mit.edu part is ingly, everyone realized that short, memorable URIs were valu-
the "domain name" of a Web server. The domain name system able commodities; the scramble for recognizable domain names,
runs on a hierarchical set of computers, which may be consulted hke candy.com and gamble.net, reached fever pitch. Speculators
to hnd out the actual Internet address (one of those numbers like began to register any name they could think of that might some-
18.23.189.58) to which packets may be sent. At the top of the day be worth more than the one-hundred-dollar registration fee.
hierarchy are five computers that store the master listâ€"and an Domain names like soap.com and sex.com were snapped up, in

126 127
w e a v i n g t h e w e b w e b o f p e o p l e



hopes of later holding out for a lucrative offer. Select names have devices in the existing domain name system that can ease the
since changed hands for large sums of money. problem. For example, if a widget company in Boston can't get
One problem is that the better domain names will wind up t n e name widget.com because it's already taken, it could try the
with the people or companies that have the most money, crip- geographically based name widget.boston.ma.us.
pling fairness and threatening universality. Furthermore, the abil- A neutral, not-for-profit organization to govern the domain-
ity to charge for a domain name, which is a scarce, irreplaceable naming process is currently being put together by the community
resource, has been given to a subcontractor, Network Solutions, at large. The original U.S.-centric nature of the domain name ser¬
which not surprisingly made profits but does not have the reputa- vice has worried some non-Americans, so any new body will
tion for accountability, or meeting its obligations. It is essential clearly have to be demonstrably international.
that domain names be primarily owned by the people as a whole, There has been a working proposal to create new top-level
and that they be governed in a fair and reasonable way by the domainsâ€"the .com or .org or .net suffixes on domain names. This
people, for the people. It is important that we not be blind to the would add top-level domains for distinct trades, such as .plastics.
need for governance where centralization does exist, just because In this way, jones.plastics and jones. electrical could be separate enti-
the general rule on the Internet is that decentralization makes ties, easing the crush a little. However, the effect would be a
central government unnecessary. repeat many times over of the ridiculous gold rush that occurred
Technically, much of the conflict is due to the mismatch for .com names, making it necessary for holders of real trade-
between the domain name structure and the rules of the social marks to protect themselves from confusion by registering not just
mechanism for dealing with ownership of names: the trademark in three domains {.com, .org, and .net) but in many more. Unless it
law. Trademark law assigns corporate names and trademarks was accompanied by a legal system for justifying the ownership of
within the scope of the physical location of businesses and the a name on some real grounds, such a scheme would hurt every-
markets in which they sell. .The trademark-law criterion of sepa- oneâ€"except those standing on the sidelines ready to make a fast
ration in location and market does not work for domain names, buck by grabbing names they never intend to use.
because the Internet crosses all geographic bounds and has no This is a relatively isolated problem with the Web, and one
concept of market area, let alone one that matches the existing the W3C has stayed almost completely clear of to date. It does
conventions in trademark law. There can be a Joe & Sons hard- serve as a good illustration of the way a single centralized point
ware company in Bangor, Maine, and a Joe & Sons fish restaurant of dependence put a wrench in the gears of an otherwise
in San Francisco. But there can only be one joeandsons.com. smoothly running decentralized system. It also shows how a
Whatever solution is found must bridge the gap between law technical decision to make a single point of reliance can be
and technology, and the chasm is fairly wide. Suppose a commer- exploited politically for power and commercially for profit,
cial entity is limited to just one domain name. Although under breaking the technology's independence from these things, and
those circumstances it might be hard to manage the persistence Weakening the Web as a universal space.
of domain names when companies changed hands, companies Even without a designed-in central point, the Web can be less
also might be prevented from snapping up names with every neutral, and more controlled, than it may seem. The Web's infra-
English word related to their area of business. There are some structure can be thought of as composed of four horizontal layers;

128 129
 w e a v i n g t h e w e b w e b oF p e o p l e


from bottom to top, they are the transmission medium, the c 0 i that happen to advertise with or make payments to the
puter hardware, the software, and the content. The transmissio 0 l
a r c h engine company. If a search engine is not giving me com-
medium connects the hardware on a person's desk, software r u ; ns
S<
letely neutral results, then I should be told about it with some
Web access and Web sites, while the Web itself is only the inf . 0r notice or icon. This is what magazines do when they run an "ar-
mation content that exists thanks to the other three layers. Thei ticle" that has been paid for by an advertiser; it is labeled "adver-
independence of these layers is important. From the softwar J
torial," or "special advertising section," or some such thing. When
engineering point of view, this is the basic principle of modular¬ companies in one layer expand or merge so they can cross layers,
ity. From the point of view of economics, it is the separation of the potential for undermining the quality of information in these
horizontal competitive markets from anticompetitive vertical, w a y s increases greatly.
integration. From the information point of view, think of editorial The trouble begins when a program that an individual
independence, the neutrality of the medium. depends on for his use of the Web, such as an operating system
The Microsoft antitrust case was big news in 1999, much of it or browser, displays an array of icons that will automatically con-
an argument about the independence in the software layer of an nect him to preferred search engines, Web sites, online programs,
operating system and a browser. In the same year, scarcely a or ISPs. Such arrangements become more troubling if a user gets
month went by without the announcement of a proposed merger a single browser/operating system that is written as one inte-
or acquisition between large companies. Two types of deals were grated software program, and cannot remove such links or nego-
taking place, the first between companies that carry data over tiate independent arrangements with other providers of similar
phone and cable T V lines, the second between content providers. services that will work with the browser/operating system.
Each of these deals was happening within one of the Web's layers. Even the hardware companies are getting into the act. In
I am more concerned about companies trying to take a verti- 1998, Compaq introduced a keyboard with four special keys: hit-
cal slice through the layers than creating a monopoly in any one ting the Search key automatically takes the user to the AltaVista
layer. A monopoly is more straightforward; people can see it and search engine. Suddenly, where a person searches the Web
feel it, and consumers and regulators can "just say no." But verti- depends on where he bought his computer. A user does not know
cal integrationâ€"for example, between the medium and content- where he stands when he hits a "Search the Web" or "Best of the
affects the quality of information, and can be more insidious. Web" button on a browser or a keyboard. These buttons or keys
Keeping the medium and the content separate is a good rule take the user into a controlled view of the world. Typically they
in most media. When I turn on the television, I don't expect it to can be set by the user to point to any search engineâ€"but few
deliberately jump to a particular channel, or to give a better pic- users change the default.
ture when I choose a channel that has the "right" commercials. I More insidiously still, it could also be possible for my ISP to
expect my television to be an impartial box. I also expect the giye me better connectivity to sites that have paid for it, and I
same neutrality of software. I want a Web browser that will show would have no way of knowing this: I might think that some
me any site, not one that keeps trying to get me to go back to its stores just seemed to have slow servers. It would be great to see
host site. When I ask a search engine to find the information it some self-regulation or even government regulation in these
can on a topic, I don't expect it to return just the sites of compa- areas.
130
131
 w e a v i n g t h e w e b w e b o f p e o p l e


The Web's universality leads to a thriving richness and d i v Happily the Web is so huge that there's no way any one com-
sity. If a company claims to give access to the world of info n y can dominate it. All the human effort people and organiza-
tion, then presents a filtered view, the Web loses its credibiljt ^ n s have put in all over the world to create Web sites and home
That is why hardware, software, and transmission companies mu ^ges is astoundingly large, and most of the effort has to do with
remain unbiased toward content. I would like to keep the conduj^ what's in the Web, not the software used to browse it. The Web's
separate from the content. I would like there always to be a choi 0
n t e n t , and thus value, will continue despite any one company's
of the unbiased way, combined carefully with the freedom to mak actions.
commercial partnerships. And when other people are making a
But consider what could happen in a year or two when
choice for me, I would like this to be made absolutely clear to me. search engines get smarter. I click the Search button on my key-,
Some might argue that bias between the layers is just the free board, or tell a search engine, "I want to buy a pair of shoes." It
market in action. But if I bought a radio and found that it supposedly heads out onto the Web to find shoe stores, but in
accessed only certain stations and not others, I'd be upset. I sup- fact brings me only to those shoe stores that have deals with
pose I could have a half dozen radios, one for each set of stations. that search engine or hardware company. The same with book-
It makes no more sense to have a half dozen computers or differ- sellers. Insurers. News. And so on. My choice of stores and ser-
ent operating systems or browsers for Web access. This is not just vices has thus been limited by the company that sells the
impractical; it fragments the Web, making it cease to be univer- computer or runs the search service. It's like having a car with
sal. I should be able to buy whichever computer, software, and a Go Shopping for Shoes button on the dashboard; when
transmission service I want and still have access to the entire pushed, it will drive only to the shoe store that has a deal with
content of the Web. the carmaker. This doesn't help me get the best pair of shoes for.
The portals represent the self-reinforcing growth of monop- the lowest price, it doesn't help the free market, and it doesn't
olies, especially those that integrate vertically. In its greater con- help democracy.
text, the battle of the portals is a battle for brand names on the
Web. It is difficult for someone to judge the quality of informa- While there are commercial incentives for vertically integrating
tion, or Web software and services, without extended experience the layers into one business, legal liability can complicate the pic-
and comparison. As a result, software or transmission companies ture. In 1998 a Bavarian court convicted Felix Somm, a former
with existing reputations can capitalize by using their names to head of the German division of CompuServe, of complicity in
attract people to their information services. The extreme would knowingly spreading pornography via the Internet. The two-year
be a company that offered transmission, hardware, software, and suspended sentence marked the first time in Germany that an
information, and then tried to brand itself as more or less equiva- online company manager had been held responsible for providing
lent to the Web. It would also be a repeat of the dial-up service access to content deemed illegal. The material was obtained from
world of A O L and CompuServe that existed before the Web, on a computers in other countries, but through CompuServe's gateway
larger scale. So far, the urge to achieve dominance has driven the to the Internet. When the boundary between the medium and the
quality on the Web upward, but any one company's attainment content is blurred, every ISP or telecommunications company is
of it would destroy the Web as we know it. in danger of being liable for content.
132
133
w e a v i n g t h e w e b w e b o f p e o p l e


Somm said he had even notified German authorities about In 1998 patrons of the Loudon County, Virginia, public
the illegal material and aided them in their investigation. Compu- library filed suit seeking to remove a filter program installed on
Serve also provided its subscribers with software they could use Internet computers at six county library branches. They claimed
to block access to offensive material. Somm may have a chance that, while the filter blocked them from accessing pornographic
for acquittal under a new German multimedia law that was sites, it also blocked them from sites with information on sex
passed after he was charged. It says that Internet service education, breast. cancer, and gay and lesbian rights. The prin-
providers can be held responsible for illegal material on their ciple here is more interesting than the bickering over details: The
servers only if they are aware of it, it is technically feasible to suit charged that the library's policy was an unconstitutional
stop it, and they do not take reasonable measures to block access form of government censorship.
to itâ€"which is what Somm and CompuServe said they did. Just how thorny these qualitative decisions can be was illus-
Somm's defense attorneys argued that no one can be aware of trated by a 1998 case described in the New York Times: "The
everything on the Internet, and that blocking access to any one American Family Association, a conservative Christian group, has
bit of it is an exercise in futility. been a vocal supporter of filtering products. So it was with some
Since the Web is universal and unbounded, there's all sorts of surprise that officials at the group recently discovered that their
junk on it. As parents, we have a duty to protect our young chil- own Web pages were being grouped with white-supremacist and
dren from seeing material that could harm them psychologically. other 'intolerant' sites blocked by a popular filter called Cyber
Filtering software can screen information under control of the Patrol. Researchers at Cyber Patrol decided, the site met the fil-
reader, to spare the reader the grief of having to read what he or ter's definition of intolerance, which includes discrimination
she deems junk. People use filters on e-mail to automatically cate- based on sexual orientation." It seems researchers had found
gorize incoming information. A n individual clearly has the per- statements on the group's page that spoke out against homosexu-
sonal right to filter anything that comes at him, just as he would ality. Cyber Patrol bans up to twelve categories of material it con-
do with regular mail: Some he opens, some he tosses into the siders inappropriate for the typical twelve-year-old, from gambling
garbage. Without this right, each day would be chaos. In the to cult sites.
future, good browsers will be able to help the user avoid links to The subjective nature of these decisions is why we set up the
Web sites that have attributes he has indicated he doesn't want to PICS system to allow anyone to customize their own objectives
have to confront, whether it's the presence of a four-letter word without imposing them on others. The key to PICS, and to any
or the fact that the site shows ads. attempt to filter, is to give the reader control, and to make differ-
But when someone imposes involuntary filters on someone ent filters available from different groups. With PICS, parents
else, that is censorship. If a library is supposed to provide a com- aren't limited to a given label provider, or even a given system of
puter that gives citizens access to the Internet, but it prevents ratings. They have a range of commercially available surveillance
access to certain types of material such as pornography, then the Programs to choose fromâ€"a choice of whom we trust.
library is deciding for the citizenry what they should be able to The larger point to remember is that laws must be written in
read. Here the library is installing itself as a central authority that relation to actions, not technology. The existing laws that address
knows better than the reader. e
gal aspects of information are sufficient. Activities such as

135
w e a v i n g t h e w e b w e b of p e o p l e


fraud and child pornography are illegal offline and online. I don't Here the liberals seem to be wanting to leverage technology
like the idea of someone else controlling the kinds of information in order to constrain government. I find it troubling when
I can access. I do believe, however, that a parent has to protect Americans of any party don't trust their political system and try
his or her child on the Internet, just as they would guard where to go around it rather than get it right. The consortium is not
their child goes physically. But the decision as to what informa- going to prevent bad laws by selectively controlling what tech-
tion adults can access needs to be up to them. nology it develops and when to release it. Technologists have to
This principle was at the center of First Amendment disputes act as responsible members of society, but they also have to cut
about the constitutionality of Internet censorship laws. When the themselves out of the loop of ruling the world. The consortium
first effort to censor the Internet was challenged in court, mem- deliberately does this. It tries to avoid acting as a central reg-
bers of the consortium felt it was important that the courts istry, a central profit taker, or a central values setter. It provides
understand how filters could act as an effective alternative to technical mechanisms, not social policies. And that's the way it
censorship. We provided background information during the will stay.
deliberations. In 1996 the United States Supreme Court over-
turned the censorship law, in part because filters enable parents The openness of the Web also means there must be a strong con-
to protect their kids without requiring the government to step in cern about business standards. Companies involved in electronic
and play nanny. But in 1998 Congress passed another censorship commerce are well aware of this, and some are making attempts
law. It's been challenged again, so that issue is far from settled. to avoid possible governmental imposition of ethical standards by
The debate has become more complex, too. Some civil liber- trying to regulate themselves, primarily with endorsements.
tarian groups claim that repressive governments could use pro- The Netcheck Commerce Bureau, for example, is a site where
grams like PICS to squelch political or social communications on companies can register their commitment to certain standards, and
the Web that the government doesn't want read: One group, the receive a corresponding endorsement. Customers can lodge
Global Internet Liberty Campaign (GILC), wrote an open letter to complaints against such companies with Netcheck. The long-
the Web Consortium saying that, to avoid this danger, W3C established U.S. Better Business Bureau has a Web site that pro-
should not release PICS Rules. PICS Rules is the part of the PICS vides similar tools. Ideally, complaints to these sites will be
technology that allows a person or group to store their prefer- monitored so that if a company doesn't do right by its customers, it
ences on a floppy disk, and give them to someone else to use. will lose the seal of approval.
G I L C was worried that the software for doing this could be mis- Some large companies are taking it upon themselves to estab-
used by repressive governments against their own people. GILC lish what is in essence a branding of quality. Since the fundamen-
also worried, according to Amy Harman of the New York Times, tal issue is determining which site to trust, if someone trusts a
that if PICS technology was widely promulgated, Congress could large company such as IBM, and I B M brands other companies as
pass a law requiring parents to adopt a particular set of PICS ethical, then the person will trust those companies, too. Indeed,
Rules. Since this would constitute government control, G I L C said IBM has developed what it calls an e-business mark, which it
the consortium should not make PICS Rules a standard. We bestows on companies it does business with that have shown a
should just bury it. commitment to delivering a secure and reliable environment for

136 J-37
w e a v i n g t h e w e b w e b oF p e o p l e


e-commerce. It's like the Underwriters Laboratory symbol or the a particular branch of alternative medicine that he believes in,
Good Housekeeping Seal of Approval. then he could use an endorsement that is based on a given jour-
Unlike regulation, endorsement can be done by anyone, of nal of alternative medicine. That's the beauty of the Web; it's a
anything, according to any criteria. This three-way independence web, not a hierarchy.
makes systems of endorsement very open. An individual can trust
a product; or an endorser, or a particular endorsement criterion. Endorsements, as a way of transmitting judgments of quality,
Self-regulation works when there is freedom to set different work easily on the Web, because they can be made with hyper-
standards and freedom of consumer choice. However, if "self- text links. However, important though this facility is, it is even
regulation" simply becomes an industry version of government, more important to understand that a link does not have to imply
managed by big business rather than by the electorate, we lose any endorsement. Free speech in hypertext implies the "right to
diversity and get a less democratic system. link," which is the very basic building unit for the whole Web.
The e-business mark may be a harbinger of the way many In hypertext, normal links are between a hypertext document
endorsements will go. People in general will not be able to figure and another external document. Embedded links are those that
out whether they trust a specific online store. So they'll resort to cause something to appear with a document; a picture appears in
"trusted" brand namesâ€"or endorsements from them. a Web page because of an embedded link between the page and
PICS was the consortium's mechanism to allow endorsements the picture. Normal hypertext links do not imply that the linked
to be coded and checked automatically. It was aimed initially at document is part of, endorsed by, or related in ownership to the
showing that a Web site meets certain criteria for lack of nudity, first document. This holds unless the language used in identifying
violence, and such. It hasn't been implemented widely because the contents of the linked document carries some such meaning.
there is no tremendous economic incentive for people to rate sites. If the creator of the first document writes, "See Fred's web page
But there may be a huge incentive when it comes to protecting [link], which is way cool," that is clearly some kind of endorse-
the privacy of personal data someone gives to an online clothing ment. If he writes, "We go into this' in more detail on our sales
store. The question is whose ratings, or settings, to trust. brochure [link]," there is an implication of common authorship. If
As a consumer, I'd like to be made aware of the endorsements he writes "Fred's message [link] was written out of malice and is a.
that have been given to a siteâ€"but without being distracted from downright lie," he is denigrating (possibly libelously) the linked
the content. Perhaps icons could appear in a window I leave open document. Clarifying the relative status of a linked document is
while I access a site, or in the border around the page I'm view- often helpful to readers, but the person has to be responsible
ing. Endorsements could be made in all fields, not just business. about what he says, just as he would in any medium.
There couAd be academic endorsements: When I'm browsing For embedded links, however, the author of the document
through research papers on heart disease, an endorsement could has responsibility, even if the contents have been imported from
appear that says a given paper has been published in a reputable another Web site, and even if the document gives the U R I for the
journal. Each reader picks the journals he trusts. An individual embedded text or image so a browser can check the original
would do the same with endorsements from associations in his source. If I write about the growth of the Web and show a graph,
profession. And if his medical association, say, happened to ignore the graph is part of my document. It is reasonable to expect me

138 J-39
w e a v i n g t h e w e b
w e b o f p e o p l e

to take responsibility for the image just as for the text. They are documents. If anything, they should be glad that more people are
logically part of the same document. Advertising embedded in a being referred to them. If someone at a meeting recommends me
site is the exception. It would be great if the H T M L distinguished as a good contact, does that person expect me to pay him for
links to "foreign" documents from links to documents with com- making reference to me? Hardly.
mon authorship, and if browsers, passed this information on to
users in some way. MYTH THREE: "Making a link to someone's publicly readable doc-
But beyond this distinction between normal and embedded ument is an infringement of privacy." The Web servers can pro-
links, certain misunderstandings still persist. Here are three vide ways to give Web site access only to authenticated people.
myths that have crept into the "common wisdom" about the Web, This technology should be used, and Web site hosting services
and my opinion as to the way hypertext protocols should be should give publishers control over access. "Security by obscu-
interpreted. rity"â€"choosing a weird U R I and not telling people about itâ€"is
not conventional, and so a very explicit agreement must be made
MYTH ONE: "A normal link is an incitement to copy the linked with anyone who is given the U R I . Once something is made pub-
document in a Way that infringes copyright." The ability to refer lic, one cannot complain about its address being passed around.
to a document (or a person or anything else) is a fundamental I do feel it is right to have protection for confidential informa-
right of free speech. Making the reference with a hypertext link tion that has.become public by accident, illegal act, or force of law
is efficient, but changes nothing else. such as a subpoena. The current assumption that once informa-
Nonetheless, in September 1998, A B C News told the story of tion has "accidentally" escaped it is free to be used is unfortunate.
a photographer who tried to sue the department store JC Penny, These are my personal feelings about how hypertext should
which had a link from its site to the Movie Database Ltd. site, be interpreted, and my intent. I am not an expert on the legalities
which in turn had a link to a Web site run by the Swedish Uni- in each country. However, if the general right to link is not
versity Network, which was said to have an illegally copied upheld for any reason, then fundamental principles of free
image of the photographer's. Fortunately, the suit was thrown speech are at stake, and something had better be changed.
out. A good default rule is that legality online is the same as it is
offline. Users, information providers, and lawyers need to reach
consensus on this. Otherwise, people will be afraid to make links
for fear of legal implications. It would soon become impossible to
even discuss things.


MYTH TWO: "Making a link to an external document makes the
first document more valuable, and therefore is something that
should be paid for." It is true that a document is made more valu-
able by links to other relevant, high-quality documents, but this
doesn't mean anything is owed to the people who created those

i/,o
141
 C H A P T E R 11




P r i v a c y




YVh en the Web started, one of the things holding it back was
often people's unwillingness to be open about their workingsâ€"
their sources and reasons behind their work. I found this frustrat-
ing myself, and would carry the banner for openness of
information while I was promoting the Web as one way of foster-
ing this openness. However, I rapidly separated the two, as the
Web does not and should not imply that all information must
always be shared. To maintain integrity, a group needs a
respected border, which in the Web is a border of information
flow. Groups need to be able to talk among themselves, and have
their own data when necessary.
Perhaps the greatest privacy concern for consumers is that,
after they have ordered enough products, companies will have
accumulated enough personal information to harm or take advan-
ce of them. With consequences ranging from the threat of junk
^ail to the denial of health insurance, the problem is serious, and
w e a v i n g t h e w e b p r i v a c y


two aspects of the Web make the worry worse. One is that infor- Internet is no one knows you're a dog." It has been followed
mation can be collected much more easily, and the other is that it recently by another cartoon in which one dog has clicked to a
can be used very easily to tailor what a person experiences. page with a picture of dog food. Because of this, the server now
To see just what can happen to my personal information, I does know it's a dog. Pretty soon the server also knows it's a
have traced how some online purveyors have used my address. dog that prefers a certain brand of dog food, elm trees, and
When I provide my address to a Web site, I put a bogus line in it, Siamese cats.
like an apartment number. Their computer regurgitates it verba- In the basic Web design, every time someone clicks on a link,
tim, so I can tell, when I get junk mail later, who has furnished their browser goes from server to server afresh, with no refer-
my address. ence to any previous transactions. The controversial tool for con-
There are more threatening scenarios. Burglars could find it sumer tracking that changes all that is the cookie. A cookie is just
very handy to know who has been buying what recently. More a code such as a reference number or account number that the
likely is the sort of abuse that occurs when a doctor divulges server assigns to the browser so as to recognize it when the same
someone's medical condition to the patient's insurance company person returns. It is much like getting an account number when
to justify the claim. Two years later, the insurance company picks opening a bank account. The cookie is automatically stored on the
the information out of its database when a prospective employer consumer's hard drive, with or without his knowledge, depending
wants to check that person's record. The person doesn't get the on his preferences.
job because of a previous medical condition and never even Most transactions between a consumer and a store involve
knows what happened. some continuity, and the cookie makes it possible to accumulate
Software can even track the pattern of clicks a person makes things in a shopping cart,, or send items to the same address as
on a Web site. If a user opens an online magazine, the publishers last time. Normally, merchants we trade with know what we
can watch which items he reads, tell which pictures he calls up have bought, and with whom we bank, and where we live, and
and in what order, and extract information about him that he we trust them. The fact that cookies are often installed on a per-
would never volunteer on a form. This is known as "click stream" son's hard drive, and talk back to the server, without any form of
information. Net Perceptions, started by a former head of permission is also valuable: It's the difference between going into
Microsoft's programming languages division, is one firm that a
store and being recognized as creditworthy, and going in and
makes software that companies can use to monitor all sorts of having to fill out identification forms all over again.
online behavior, from the amount of time a visitor spends reading However, some commentators see cookies as entirely evil. By
about a product to what pages they print on their printer. de
fault, most browsers accept all cookies automatically, but then
If an advertiser runs ads on different sites and finds a per- again most also offer the user the option of prompting them with
son's click stream on a certain selection of the sites, it can build a n
alert notice before the computer accepts a cookie, or of simply
up an accurate profile of sites that person visits. This information refusing it. The problem is not in the cookie itself, over which
can then be sold to direct marketers, or whomever. A famous car- t h e
user has control. The problem is that there is no knowing
toon drawn early in the Internet's life depicts two dogs sitting a • f w
hat information the server will collect, and how it will use that
a computer. One explains to the other, "The great thing about the Inf
orniation. Without that information the user can make choices

144
 w e a v i n g t h e w e b p r i v a c y


based only on fear and doubt: not a stable basis for building ¡ The next step is to make it possible for my browser to do this
ety on the Web. j o r m e _ o t just to check, but to nègotiate for a different privacy
n



A Web site also can change chameleon-like according to < policy one that will be the basis for any subsequent release of
is looking at it, as if it were a brochure being printed for that information. With privacy software, a Web site provider and
person. Imagine an individual visiting the Web page of a political browser can do just that.
candidate, or a controversial company. With a quick check of that Consider a company selling clothing over the Internet. It
person's record, the politician or company can serve up just the might declare its privacy policy as follows:* "We collect your
. right mix of propaganda that will warm that particular person's name, age, and gender to customize our catalogue pages for the
heartâ€"and tactfully suppress points he or she might object to. I s type of clothing you are likely to be interested in and for our own
this just effective targeted marketing, or deception? It depends on product development. We do not provide this information to any-
whether we know it is happening. one outside our organization. We also collect your shipping infor-
Europe has tried to solve part of this problem with strict reg- mation. We may distribute this information to others."
ulation. European companies have to keep secure the information For these things to be negotiated automatically, the prefer-
they hold on customers, and are barred from combining data- ences set by a user and the privacy policy have to be set up in
bases in ways that are currently quite legal in the United States. machine-readable form using some common set of categories for
Consumers in Europe also have the right to look at and correct different sorts of data and different ways of using it.
databases that contain information about them. In the United The World Wide Web Consortium is creating a technology
States, laws that protect consumers from having their information that will allow automatic negotiation between a user's browser
resold or given away are very weak. The government has hoped and a store's server, leading to an agreement about privacy. The
that some sort of self-regulation will come into force. Platform for Privacy Preferences Project (P3P) will give a com-
The good news is that the Web can help. I believe that the puter a way of describing its owner's privacy preferences and
privacy I require of information I give away is something I ought demands, and give servers a way of describing their privacy poli-
to have choice about. People should be able to surf the Web cies, all done so that the machines can understand each other and
anonymously, or as a well-dehned entity, and should be able to negotiate any differences without a person at either end getting
control the difference between the two. I would like to be able involved.
to decide who I will allow to use my personal information and I believe that when a site has no privacy policy there ought to
for what. be a legally enforced default privacy policy that is very protective
Currently, a responsible Web site will have a privacy policy of the individual. Perhaps this view shows my European roots.
one click from the bottom of the home page. One site might sell And it may sound counter to my normal minimalist tendencies.
any information it gets to direct-mail firms or advertisers. But lack of such enforcement allows a company to make what-
Another may record every page a visitor views. Another might ever use it can of whatever private data it can somehow extract.
not distribute any information under any circumstances. I could In 1998 the Federal Trade Commission did a survey of Web
read this carefully and decide whether to proceed, but in practice sites and found that very few had a privacy policy, including sites
I usually don't have time to read it before rushing in. that took information from children. The findings were so dramatic
146 . M7
 p r i v a c y
w e a v i n g t h e w e b

this problem. And it should not be up to the consortium or any
that President Clinton called a two-day Internet privacy meeting
other technical body to solve it.
in Washington with industry and government officials. The
Perhaps the most notorious violation of privacy over the Web
results also prompted the Federal Trade Commission to consider
was the sudden release late in 1998 of details from the U.S. Inde-
regulating privacy policies.
pendent Council's report about President Clinton's sexual activi-
As is so often the case, the possibility of regulation has
ties. This information was purposely exposed to millions of
prompted industry to make some moves toward self-regulation. In
people, contrary to many people's concepts of respect for the indi-
June 1998, Christine Varney, a former F T C commissioner, put
vidual or family. We can use the power of the Web to connect any-
together a group of about fifty companies and trade groups called
thing and everything to great effect, or to do devastating damage.
the Online Privacy Alliance. Members included A O L , AT&T,
Episodes like this help us recognize how rapidly the widespread
Microsoft, Netscape, the Direct Marketing Association, and the
distribution of information could cripple our societyâ€"and each of
U.S. Chamber of Commerce. They said they would clearly reveal
us personallyâ€"if absolutely all information remains public.
what information they collected on all their various Web sites and
how it would be used. They also said they would give consumers
some choice about how personal data could be used, including the No one will take part in the new weblike way of working if they
ability to not allow their information to be sold to third parties. do not feel certain that private information will stay private. In a
The Better Business Bureau Online is also addressing the matter group, they will also remain oh the sidelines if they feel that
with an endorsement serviceâ€"a privacy seal it will grant to wor- what they say or write will not remain confidential, or if they
thy Web sites. The program features privacy-standard setting, ver- can't be sure of whom they are communicating with.
ification, monitoring, and review of complaints. Public key cryptography (PKC) offers one way to achieve the

Some regulators maintain that since there is no mechanism four basic aspects of security: authenticity, confidentiality, integrity

for enforcement, this kind of effort does not go far enough. of messages, and nonrepudiatability. Each person has a number

Tighter control, they say, is needed, especially when it comes to that everyone knows (the public key), and another, related number

protecting information about children. They maintain that any that no one else ever has (the private key). Devised more than two

abuse of information about adults or children should be illegal. decades ago, P K C provides a form of encryption in which an out-

But the Online Privacy Alliance is a good start, at least in creat- going message is scrambled according to the receiver's public key.

ing a system of endorsements, which will cause more consumers The scrambled message can then be decoded only by a receiver

to gravitate toward sites that comply. This will put pressure on who has the unique matching private key to unlock it. A leading

others to do the same. Ideally, such groups will set privacy prac- form of public key cryptography is RSA, named after its developers,

tices that will be automatically checkable with P3P. Ron Rivest, Adi Shamir, and Leonard Adleman, all of whom were
a
t MIT's Laboratory for Computer Science in 1977 when they
Of course, any privacy negotiation is only as trustworthy as
invented it.
the site's, proprietor. However, if a company has, through its Web
server, made an undertaking to preserve privacy, and broken that Deducing whether someone or someplace is authentic begins
undertaking, then it has acted fraudulently. There are conven-
W l
* h common sense. If a Web site offers a deal that seems too

tional laws to deal with this transgression. Software can't solve S°od to be true, it probably is. Tougher, however, is figuring out

149
148
w e a v i n g t h e w e b p r i v a c y


whether the Web site of a well-known clothing store is indeed ^ d virtually impossible to crackâ€"so impossible, in fact, that
operated by that store. Anyone can make a site that looks like a since its development more than twenty years ago the U.S. gov-
clothing store. Crooks could even have an elaborate impostor site ernment has blocked the export of strong cryptography by classi-
that takes an order, passes it to the real store, sends the store's fying it as "munition." Some other governments have reacted in
communication back, and in the meantime steals the credit-card similar ways, blocking export, or banning its use, for fear that ter-
number. And unlike a physical facade, the fake store will look and rorist groups will be able to communicate without government
feel indistinguishable from the real one. Currently there is some being able to tap into their conversations.
attempt to make the domain name system more secure, but at the The counterargument points to George Orwell's vision in his
moment authenticity relies mainly on the security from intrusion book 1984, in which the National Security Agency becomes Big
of the domain servers (which tell the browser where, for example, Brother, able to monitor a person's every move. It argues that
www.acme.com is. on the Internet) and the connections between without the basic right of the citizen to discuss what he or she
them. Public key authentication would be much better. wants, the people are left at the mercy of potential dictatorial ten-
Confidentiality consists in knowing that no one else can dencies in government.
access the contents of a communication. Once again, criminals or The balance in governmental power is always a tricky thing.
spies can intercept a communication to a clothing store and skim But the debate is almost moot in this case, because encryption
off credit-card numbers being sent electronically, or eavesdrop on technology has been written in many countries of the free world.
supposedly private conversations between people in a group. The U.S. export ban frustrates people who simply want, say, to
Encryption technology prevents this by scrambling the messages. buy clothes from another country. It infuriates software manu-
Anyone browsing a site whose U R I starts with https: is using an factures who have to make two versions of each product, one
encryption technology called Secure Socket Layer. Normally, with strong P K C and the other with a specifically weakened ver-
however, cryptography is used only to make sure no one except sion for export, and then devise ways of trying to prevent the
the server can read the communicationâ€"not to verify that the strong one from crossing borders. It hobbles the Open Source
server is really who it says it is. community, in which distribution of the source code (original
The integrity of messages involves making sure no one can written form) of programs is a basic tenet. To ridicule the export
alter a message on the Internet without being detected, and non- law, PKC programs have been printed on T-shirts, and in machine-
repudiatability means that if I have sent a message, I can't later readable fonts in booksâ€"which cannot be subject to export
maintain that I did not. PKC provides technology to assure these, controls.
too. If I use the software to add to a message I send (or a Web There is another reason why PKC has not been adopted: It can
page I write), a number at the end called a digital signature allows only be used in conjunction with a system for telling your com-
the receiver to verify that it was I who sent it and that it has not puter which public keys to trust for which sorts of things. This is,
been tampered with. The consortium has a project for apply & in
°f course, a very important but also difficult thing to express. An
digital signatures to documents. uidividual's ability to express trust is essential, because without
If P K C is so well understood, why are we not using it? O n e t r i a t
trust many uses of the Web, from collaborative work to elec-
reason is the government's fear of loss of control. It is easy to use, tronic commerce, will be socially impossible.

150
w e a v i n g t h e w e b p r i v a c y


Authenticity and confidentiality are not problems new to the- All these certificate authorities will vouch for the identity of people
Web. They have been solved, in principle, for electronic mail^ a nd their keys. They generally sell certificates, which expire after
Pretty Good Privacy (PGP) , and Secure M I M E are two standards? a certain number of months. But I don't see a button to set myself
for digitally signing mail (to authenticate the person who sent it}' u p to issue a certificate to a friend or relative whom I also trust.
and encrypting it (to stop anyone else from reading it). pGP would allow this.
PGP is more or less a grassroots system. It is a web of trust,: The Web worked only because the ability of anyone to make
An alternative, Public Key Infrastructure (PKI), is basically a tree- a link allowed it to represent information and Telationships how-
like way of doing things. In either PGP or PKI, a user's computer^ ever they existed in real life. The reason cryptography is not in
associates a key with a person by holding a hie called a certifi- constant use in representing trust on the Web is that there is not,
cate. It typically carries the person's name, coordinates, and pub- yet, a weblike, decentralized infrastructure.
lic key. The certificate itself is digitally signed with another
public key of someone else the user trusts. He knows that it is The PGP system relied on electronic mail, and assumed that
the other person's key because it says so on a certificate that was everyone held copies of certificates on their hard disks. There
signed with a key of a different party he trusts. And so on, in a were no hypertext links that allowed someone to point to a cer-
chain. tificate on the Web. Clearly, it should be much easier to introduce
The social structure assumed by PGP is that chains of trust a Web of Trust given the Web.
will be made through anyoneâ€"a person's family, friends, college, I mentioned that both PGP and PKI made two assumptions:
employer. If an individual was authenticating a message from a that we trust a person, and that if we do, we just have to link a
colleague, he probably would use a certificate signed by their person with a key. Many pointless arguments and stalling points
common employer. That is the path of trust. have involved exactly what constitutes a person, and how to
The PKI system, planned by industry to enable electronic establish the identity of a person. In fact, in most situations it
commerce, assumes that people trust just a few basic "roots" does not matter who the person "is" in any unique and funda-
from which all authority flows. A few certificate authorities dele- mental way. An individual is just interested in the role the person
gate the right to issue certificates to their commercial partners. plays, which is represented by a public key. All we need to do is
They in turn can delegate the right to issue certificates to other, find a language for talking about what can be done with different
smaller authorities. There is a tree, and money and authority flow keys, and we will have a technical infrastructure for a Web of
up and down it. Trust. If we play our cards right, the work at the consortium in
Browsers are now slowly being equipped to work with the languages for the Web (which I will describe in chapter 13) will
Public Key Infrastructure. If I open the browser preferences on end up producing a Web of Trust. Then the Web and the Web of
my Internet Explorer now, I see that I can chose to accept certifi- Trust will be the same: a web of documents, some digitally
cates signed by Microsoft, ATT, G T E , M C I , Keywitness Canada Sl
gned, and linked, and completely decentralized. The consortium
Inc., Thawte, and Verisign. In the equivalent list in Netscape, I w
u l not seek a central or controlling role in the Web of Trust; it
see ATT, BBN, BelSign, Canada Post, Certisign, G T E , GTIS, IBM, W l
U just help the community create a common language for
Integrion, Keywitness, M C I Mall, Thawte, Uptime, and Verisign• expressing trust.

152 • 153
w e a v i n g t h e w e b


The Web of Trust is an essential model for how we really
work as people. Each of us builds our own web of trust as we
mature from infancy. As we decide what we are going to link to,
read, or buy on the Web, an element of our decision is how much
we trust the information we're viewing. Can we trust its pub-
lisher's name, privacy practices, political motivations? Sometimes
we learn what not to trust the hard way, but more often we
inherit trust from someone elseâ€"a friend or teacher or family
memberâ€"or from published recommendations or endorsements
by third parties such as our bank or doctor. The result of all this
activity creates a web of trust in our slice of society.
Automated systems will arise so negotiations and transac-
tions can be based on our stated criteria for trust. Once we have
these tools, we will be able to ask the computer not just for
information, but why we should believe it. Imagine an O h Yeah?
button on a browser. There I am, looking at a fantastic deal that
can be mine just for the entry of a credit-card number and the
click of a button. I press the Oh Yeah? button. My browser chal-
lenges the server to provide some credentials. Perhaps this is a
list of documents with digitally signed endorsements from, say,
the company's bank and supplier, with the keys to verify them.
My browser rummages through these with the server, looking to
be convinced that the deal is trustworthy. If it's satisfied, good
for me, I got a deal. If not, I probably just saved myself some
grief.
It would be wrong to assume that the Web of Trust is impor-
tant primarily for electronic commerce, as if security mattered
only where money is concerned. The Web is needed to support
all sorts of relationships, on all levels, from the personal, through
groups of all sizes, to the global population. When we are work-
ing in a group, we share things we would not share outside that
group, like half-baked ideas and sensitive information. We do so
because we trust the people in the group, and trust that they
won't divulge this information to others. To date, it has been diffi'

i5<t
 C H A P T E R 12




M i n d to M i n d




I have a dream for the Web . . . and it has two parts.
In the first part, the Web becomes a much more powerful
means for collaboration between people. I have always imagined
the information space as something to which everyone has imme-
diate and intuitive access, and not just to browse, but to create.
The initial WorldWideWeb program opened with an almost blank
page, ready for the jottings of the user. Robert Cailliau and I had
a great time with it, not because we were looking at a lot of stuff,
but because we were writing and sharing our ideas. Furthermore,
the dream of people-to-people communication through shared
knowledge must be possible for groups of all sizes, interacting
electronically with as much ease as they do now in person.
In the second part of the dream, collaborations extend to
computers. Machines become capable of analyzing all the data on
the Webâ€"the content, links, and transactions between people and
computers. A "Semantic Web," which should make this possible,

157
w e a v i n g t h e w e b m i n d t o m i n d


has yet to emerge, but when it does, the day-to-day mechanisms An essential goal for the telecommunications industry (and
of trade, bureaucracy, and our daily lives will be handled by regulatory authorities) should be connecting everyone with per-
machines talking to machines, leaving humans to provide the inspi- manent access. The problem till now has not been technology, but
ration and intuition. The intelligent "agents" people have touted rather regulations that control what telephone companies can
for ages will finally materialize. This machine-understandable charge for access, and the lack of agreement about how other
Web will come about through the implementation of a series of companies that might want to provide Internet access can lease
technical advances and social agreements that are now beginning the copper wire that goes to every home. With some wiser regu-
(and which I describe in the next chapter). lation, in some cases spurred on by competition from cable com-
Once the two-part dream is reached, the Web will be a place panies that lay their own cables to people's doors, before too long
where the whim of a human being and the reasoning of a I should be able to walk up to a screen, see it quickly glow with
machine coexist in an ideal, powerful mixture. my home page on it, and follow a link immediately. This simple
Realizing the dream will require a lot of nitty-gritty work. difference in timing will dramatically change the way we use
The Web is far from "done." It is in only a jumbled state of con- computers, making the experience more like getting out a pen
struction, and no matter how grand the dream, it has to be engi- rather than getting out a lawnmower. Computers will be there
neered piece by piece, with many of the pieces far from when we suddenly have an idea, allowing us to capture it and
glamorous. preventing the world from losing it.
It is much easier to imagine and understand a more enlight- Let's clear our minds about what we will see on these won-
ened, powerful Web if we break free of some of the world's cur- derful new computers. Today there is a desktop with Various
rent assumptions about how we use computers. When I want to folders and "applications." One of these applications is a Web
interact with a computer, I have to wait several minutes after browser. In this scheme, my entire screen is taken up by my
turning it on before it is ready to converse. This is absurd. These local computer, while all the information in the rest of the acces-
machines are supposed to be there for us, not the other way sible world is relegated to a small area or icon within it. This is
around. So let's begin our thinking about a new world by imagin- inside out.
ing one in which a computer screen is available whenever we The job of computers and networks is to get out of the way,
want it. to not be seen. This means that the appearance of the informa-
In the same spirit, we should jettison our assumptions about tion and the tools one uses to access it should be independent of
Internet access. Why should we have to wait while a computer where the information is storedâ€"the concept of location indepen-
connects to the Internet by making a phone call? The Internet dence. Whether they are hypertext pages or folders, both valid
isn't designed to be like that. It is made so that, at any time, a little genres of information management, they should look and feel the
postcardlike packet of a few hundred characters could be dropped same wherever they physically happen to be. Filenames should
into it by one computer, and in a fraction of a second be at its des- disappear; they should become merely another form of U R I .
tination on the other side of the world. That is why clicking on an T
ben people should cease to be aware of URIs, seeing only
icon can take us very quickly to a Web site. The bother of having hypertext links. The technology should be transparent, so we
to make a phone call wrecks the idea of instant availability. lr
iteract with it intuitively.

158 159
w e a v i n g t h e w e b m i n d t o m i n d


The next step would be protocol independence. Right now information. For example, the minutes of a meeting may contain
every time I write something with a computer, I have to choose "action item." X M L allows the person taking the minutes to
whether to open the "electronic mail" application or the "net make a new document type that includes < action > as a new tag.
news" application or the "Web editor" application. The mail If the minutes are recorded in H T M L , this might be lost, because
news, and Web systems use different protocols between comput- HTML's general set of tags don't include < action > and the person
ers, and effectively, I am being asked to select which protocol to taking the minutes can't create one. A n X M L document is typi-
use. The computer should figure this out by itself. cally richer: The information it contains is rnore well defined.
Location independence and protocol independence would be This will allow such things as spreadsheets, calendar files,
very simple if all the software on a computer were being rewrit- e-mail address books, and bank statements that have not used
ten from scratch. Unfortunately, it isn't. The required change to interoperable standard formats to have them developed quickly,
the modular design of operating systems and applications would dramatically increasing the interoperability in, for example, typi-
be significant. Indeed, whether or not the terms operating system cal office documents. This is the primary excitement behind the
and application would survive is not clear. But since software XML revolution: avoiding the information lost when such docu-
engineers are very inventive, and the stakesâ€"an intuitive inter- ments are translated into H T M L and thereby lose their ability to
faceâ€"are high, I am optimistic. be understood as spreadsheets, calendars, bank statements, or
As we look at the way a-person uses the Web, it is simplest to whatever.
improve the reception of information by adding new forms of The threat is that when a company introduces a new docu-
graphics and multimedia. It is more difficult to imagine how best ment type, no one else will understand it. X M L makes it easy for
to allow a person to interact with the information, to create and everyone to create their own tags or entire markup languages.
modify it. Harder still is imagining how this computer screen can We might therefore see an end to the idyllic situation that has
be used to allow one person to interact as one of many people prevailed thus far on the Webâ€"the predominance of H T M L ,
interacting as a group. This is the order in which development which has helped all of us share documents easily. Can it be
has occurred to date, and will occur in the future. that, a decade into the Web's existence, X M L will give us a free-
The X M L revolution, mentioned in chapter 9, that has taken dom that forcibly leads us back toward myriad incompatible lan-
place over the last few years and is now reaching the mainstream guages? This is indeed a serious possibility, but one that has
has provided a solid foundation for much of the new design been anticipated.
inside and outside the consortium. Even though the computer The extensible X in XML means anyone can invent new tags,
markup languages for hypertext and graphics are designed for but they can't add them to someone else's tags. An X M L docu-
presenting text and images to people, and data languages are ment can be made of a mixture of tags from more than one name-
designed to be processed by machines, they share a need for a space, but each namespace is identified by a U R L Thus any X M L
common, structured format. X M L is it. document is completely defined using the Web. This is a huge
X M L is both a boon and a threat to the Web dream. The great step forward from the old H T M L days in which anyone could
thing is that it stems the tide of information loss. It allows anyone make up their own version of what < table > meant, for example,
to create any kind of tag that can capture the intent of a piece of with no ambiguity. The X M L namespaces change the rules of
160 161
w e a v i n g t h e w e b m i n d t o m i n d


technology evolution by making every step, whether open or pr . 0 can bring inside their walls. Although it takes a little work to set
prietary, well defined. up the access control for a corporate or family intranet, once that
It is important to remember that X M L does not replace has been done the Web's usefulness is accelerated, because the
H T M L . It replaces the underlying S G M L on which H T M L was participants share a level of trust. This encourages more sponta-
built. H T M L can now be written as X M L . In fact, it is possible t 0
neous and direct communication.
create a valid X M L document that will also work with old HTML To be able to really work together on the Web, we need much
browsers. (The specification for doing this is X H T M L . ) better tools: better formats for presenting information to the user;
more intuitive interfaces for editing and changing information;
When I proposed the Web in 1989, the driving force I had in seamless integration of other tools, such as chat rooms, and
mind was communication through shared knowledge, and the audio- and videoconferencing, with Web editing. We need the
driving "market" for it was collaboration among people at work • ability to store on one server an annotation about a Web page on
and at home. By building a hypertext Web, a group of people of another; simple access controls for group membership, and for
whatever size could easily express themselves, quickly acquire tracking changes to documents. While some of this work involves
and convey knowledge, overcome misunderstandings, and reduce leading-edge research, a lot of it consists of trying to adapt exist-
duplication of effort. This would give people in a group a new ing computer systems to the global hypertext world.
power to build something together.
People would also have a running model of their plans and For people to share knowledge, the Web must be a universal space
reasoning. A web of knowledge linked through hypertext would across which all hypertext links can travel. I spend a good deal of
contain a snapshot of their shared understanding. When new my life defending this core property in one way or another.
people joined a group they would have the legacy of decisions Universality must exist along several dimensions. To start
and reasons available for inspection. When people left the group with, we must be able to interlink any documentsâ€"from drafts
their work would already have been captured and integrated. As to highly polished works. Information is often lost within an
an exciting bonus, machine analysis of the web of knowledge organization when a "final document" of some kind is created at
could perhaps allow the participants to draw conclusions about the end of an endeavor. Often, everything from the minutes of
management and organization of their collective activity that they meetings to background research vanishes, and the reasoning
would not otherwise have elucidated. that brought the group to its endpoint is lost. It might actually
The intention was that the Web be used as a personal infor- still exist on some disk somewhere, but it is effectively useless
mation system, and a group tool on all scales, from the team oí because the finished document doesn't link to it. What's more,
two creating a flyer for the local elementary school play to the different social and practical systems isolate documents of differ-
world population deciding on ecological issues. ent levels from each other: We don't insert random notes into fin-
I also wanted the Web to be used just as much "internally" a s
. ls
hed books, but why not, if they are relevant and insightful? At
externally. Even though most of the first ten servers, like the one the consortium today, no one can mention a document in a meet-

at C E R N or SLAC, would be called intranet servers today, organi-
m
g unless they can give a U R I for it. Our policy is "If it isn't on
r
he Web, it doesn't exist," and the cry often heard when a new
zations and families are just beginning to see the power the
162 163
w e a v i n g t h e w e b m i n d t o m i n d


idea is presented is "Stick it in Team Space!"â€"a directory for con- from these different areas. By connecting across groups, people
fidentially saving documents not otherwise on the Web. All mail also provide organization and consistency to the world. It is
is instantly archived to the Web with a persistent U R I . it \ s u nusual for an individual to support environmental policies on a
already hard to imagine how it could have been any other way global level but then plan to dump chemicals into the local river.
The Web of work and play must be able to intertwine half-baked My original vision for a universal Web was as an armchair aid
and fully baked ideas, and Web technology must support this. to help people do things in the web of real life. It would be a mir-
Another dimension critical to universality is the ability to link ror, reflecting reports or conversations or art and mapping social
local material to global. When an endeavor is put together that interactions. But more and more, the mirror model is wrong,
involves groups of different scalesâ€"whether a software engineer- because interaction is taking place primarily on the Web. People
ing project such as mine at C E R N , or an elementary school edu- are using the Web to build things they have not built or written or
cation project that is part of a town initiative and uses federal drawn or. communicated anywhere else. As the Web becomes a
fundsâ€"information has to come from many levels and has to be primary space for much activity, we have to be careful that it
cross-linked. allows for a just and fair society. The Web must allow equal
Similarly, universality must exist across the spectrum of cost access to those in different economic and political situations;
and intention. People and organizations have different motiva- those who have physical or cognitive disabilities; those of differ-
tions for putting things on the Web: for their own benefit, com- ent cultures; and those who use different languages with different
mercial gain, the good of society, or whatever. For an information characters that read in different directions across a page.
system to be universal, it can't discriminate between these. The
Web r.iust include information that is free, very expensive, and The simplest factor controlling the Web as a medium for commu-
every level in between. It must allow all the different interest nication between people is the power of the data formats used to
groups to put together all manner of pricing and licensing and represent hypertext, graphics, and other media. Under pressure
incentive systems . . . and always, of course, allow the user to because of their direct visibility and impact on the user's experi-
"just say no." ence, these have advanced relatively rapidly, because each
The reason we need universality on all these levels is that medium has been essentially independent of the others.
that's how people operate in the real world. If the World Wide One might have expected that graphics formats would have
Web is to represent and support the web of life, it has to enable been standardized long ago, but the Web introduced new stresses
us to operate in different ways with different groups of different that are forcing quite an evolution. Marc Andreessen gave
sizes and scopes at different places every day: our homes, offices, browsers the ability to display graphics right inside a document,
schools, churches, towns, states, countries, and cultures. It must mstead of relegating them to a separate window. He happened to
also transcend levels, because creative people are always crossing Pick the Graphic Interchange Format (GIF) defined by Compu-
boundaries. That is how we solve problems and innovate. Serve. Soon, people also started using the standard J P E G (Joint
Information must be able to cross social boundaries, too. O u r
Photographic Experts Group) format for photographs. These two
family life is influenced by work. Our existence in one group formats reigned supreme until Unisys announced that it had ended
affects that in another. Values and actions are fed by all the ideas U
P being the owner of a patent on the compression technology

164 165
w e a v i n g t h e w e b m i n d t o m i n d


used to make G I F images and that they would be charging license and video, were if nothing else a launch for SMIL. The language
fees. A small group of enthusiasts proposed an alternative, Portable c a n also effectively save bandwidth. Often a T V signalâ€"say, a
Network Graphics (PNG), based on an open compression technol- news broadcastâ€"has a talking head that takes up maybe a quar-
ogy, and generally superior to GIF. The consortium members ter of the screen, a still image or map in the background, and per-
agreed to endorse P N G as a W3C recommendation. haps a caption, not to mention basketball scores scrolling across
The recent moves to put the Web on everything from televi- the bottom of the screen. Transmitting all that as video data takes
sions to mobile phone screens have made the need for device a lot of bandwidth. SMIL allows the relatively small amount of
dependence very clear. This has prompted even newer graphics data about images that are actually moving to be sent as video,
formats that are more capable of displaying an image on screens and integrated with the still images that are transmitted to the
of different sizes and technologies. Both J P E G and P N G describe a viewer's screen in ways that require much less bandwidth.
picture in terms of the square grid of pixels that make up a com-. Running through all the work on hypertext, graphics, and
puter screen. The consortium is developing a new format for multimedia languages are concerns about access for all, indepen-
drawings that will describe them as abstract shapes, leaving the dent of culture, language, and disability. The consortium's Web
browser free to fill in the pixels in such as way that the image can Accessibility Initiative brings together people from industry, dis-
be shown with optimal clarity on a wristwatch or a drive-in movie ability organizations, government, and research labs to devise pro-
screen. The format, called scalable vector graphics, is based on tocols and software that can make the Web accessible to people
X M L . It will also dramatically speed up the delivery of documents with visual, hearing, physical, and cognitive or neurological dis-
containing drawings, which will open the door to all sorts of new abilities. The work ranges widely, from review of W3C technolo-
ways of interacting between a person and a Web site. And because gies to ensure that they support accessibility to development of
it is in X M L , it will be easy for beginners to read and write. We accessibility guidelines for Web sites, browsers, and authoring
may soon see all kinds of simple animated graphical interfaces. tools, and development of tools to evaluate accessibility. Much of
Virtual Reality Modeling Language (VRML) is another pillar, this works only when those building Web sites have taken a little
being created for three-dimensional scenes. I expected 3D to really care about how they have done it. The disability and technical
take off, and still don't quite understand why it hasn't. Sending the communities got together to produce a set of guidelines about the
details of a 3D scene takes relatively few bytes compared, for most effective and practical steps to take: recommended reading
example, with video. It does require the user to have a fast com- for webmasters.
puter, to manipulate the scene as the user moves around it. Per- The consortium also has an internationalization activity that
haps the power of the average processor just isn't high enough yet- checks that new specifications will work in different alphabets,
Integrating many different text, image, audio, and video whether they are Eastern or Western, read right to left, left to right,
media into one Web page or show will be greatly helped by the 0 r U
P and down. Conversions can get complicated, but the com-
Synchronized Multimedia Integration Language (SMIL; "smile )• puter industry is making energetic efforts to extend operating sys-
SMIL will make seamless coordination simple, even for author tems to support the display of all kinds of written scripts, including
with limited Web design experience. The notorious Clinton tap »- eS
^ abic, Hindi, Korean, Chinese, Japanese, Thai, and Hebrew.
r




relayed over the Web in windows with mixtures of graphics, HTML 4.0 already provides a number of internationalization
166
167
w e a v i n g t h e w e b m i n d t o m i n d


features, including the ability to mark text as to which language the T V producer has decided we should see next. But my defini-
it is in, and to order text from right to left. tion of interactive includes not just the ability to choose, but also
The primary principle behind device independence, and the ability to create. We ought to be able not only to find any
accessibility, is the separation of form from content. When the sig- kind of document on the Web, but also to create any kind of doc-
nificance of a document is stored separately from the way it ument, easily. We should be able not only to follow links, but to
should be displayed, device independence and accessibility create themâ€"between all sorts of media. We should be able not
become much easier to maintain. Much of this is achieved with a only to interact with other people, but to create with other people.
sfyZe sheetâ€"a set of instructions on how to present or transform a Intercreativity is the process of making, things" or solving problems
printed page. Hakon Lie, who worked with me at C E R N and later together. If interactivity is not just sitting there passively in front
at the consortium, led the development of Cascading Style Sheets of a display screen, then intercreativity is not just sitting there in
(CSS) to make this possible. A new, related language with differ- front of something "interactive."

ent capabilities, XSL, is also in the works. There is even an With all this work in the presentation of content, we still
"aural" style-sheet language, part of CSS2, to explain to a browser have really addressed only the reading of information, not the
how a Web page should sound. writing of it. There is little to help the Web be used as a collabo-
The growing list of graphics formats relate primarily to static rative meeting place. Realizing this early on, the consortium held
displays. But some people feel a Web page isn't sufficiently excit- a workshop to find out what was needed. The result was a long
ing unless it moves. At a minimum, they want the page to change shopping list of capabilities, things like strong authentication of
as a user interacts with it. Pop-up balloons and menus, and forms group members, good hypertext editors, annotation systems (sim-
that fill themselves in, are simple examples we find today on the ilar to the little yellow paper sticky notes), and tools for proce-
Web. These work because a small program, or script, is loaded dures such as online voting and review.
with the page. It operates the page like the hand inside a puppet, Some of the results have been satisfying. SMIL was one, inte-
in response to the user's actions. This has created a crisis in inter- grating various media and possibly allowing a real-time collabora-
operability, however, because the connection between the script tive environment, a virtual meeting room, to be constructed.
and the Web page, the hand and the puppet, is not standard for Others are still in the wings. A long-standing goal of mine had
different kinds of style sheets. To fix this, the consortium is work- been to find an intuitive browser that also, like my WorldWideWeb,
ing on a Document Object Model (DOM), a set of standards for allows editing. A few such browser/editors had been made, such as
that interface. Unfortunately, it is much more difficult to make AOLpress, but none were currently supported as commercial prod-
these animated pages accessible to voice browsers and screen ucts. Few items on the wish list for collaborative tools had been
readers. O n the positive side, the D O M interface should provide achieved. At the consortium we wondered what was wrong. Did
a powerful way for accessibility tools such as document readers People not want these tools? Were developers unable to visualize
to access the document structure within a browser. them? Why had years of preaching and spec writing and encour-
agement got hardly anywhere?
The media may portray the Web as a wonderful, interactive place 1
became more and more convinced that the only way to find
where we have limitless choice because don't have to take what ° U t W h a t w
a s holding back the development of collaborative tools

168 169
w e a v i n g t h e w e b m i n d t o m i n d


was. to try to develop them ourselves. Our policy had always u p a list of all versions, with details about who made which
been that we would use whatever commercial tools were avail¬ changes when, and revert to an older version if necessary. This
able to get our own work done. At a consortium team retreat in provides everyone with a feeling of safety, and they are more
Cambridge, I suggested we start trying all the experimental solu- inclined to share the editing of a piece of work. Jigsaw and
tions being tinkered with in the community, and even develop Amaya allow our team space to come alive as our common room,
them further. Perhaps then we would stumble upon the real internal library, and virtual coffee machine around which staff
problems, showing the way toward solutions. members who are in France, Massachusetts, Japan, or on an air-
We concluded that to do this, we needed a nucleus of people plane can gather. •
who would try various new collaboration technologies, just to see Making collaboration work is a challenge. It is also fun,
what happened. They would help the entire consortium staff because it involves the most grassroots and collegial side of the
become early adopters of experimental software. This new policy, Web community. All Web code, since my hrst release in 1991,
which we called Live Early Adoption and Demonstration (not has been open source software: Anyone can scoop up the source
coincidentally, L E A D ) , meant that we entitled ourselves to eat codeâ€"the lines of programmingâ€"and edit and rebuild them, for
our own dog food, as far as our very limited resources would free. The members of the original www-talk mailing routinely
allow. It meant that we'd be testing new protocols not on their picked up new versions of the original Web code library "lib-
own, but in the context of our actual, daily work. It also meant www." This software still exists on the consortium's public server,
that, with only a handful of programmers, we would be trying to www.w3.org, maintained for many years by Henrik Nielsen, the
maintain the reliability of these experimental products at a level cheerful Dane who managed it at C E R N and now MIT. Libwww
high enough to allow us to actually use them! is used as part of Amaya, and the rest of Amaya and Jigsaw are
We are only in the early stages, but we now have an environ- open source in the same way. There are a lot of people who may
ment in which people who are collaborating with the consortium not be inclined to join working groups and edit specifications, but
write and edit hypertext, and save the results back to our server. are happy to join in making a good bit of software better. Those
Amaya, the browser/editor, handles H T M L , X M L , Cascading who are inspired to try Amaya or Jigsaw, want to help improve
Stylesheets, Portable Network Graphics, and a prototype of Scal- them, develop a product based on them, or pick apart the code
able Vector Graphics and Math M L . While we have always devel- and create an altogether better client or server can simply go to
oped Amaya on the Linux operating system, the Amaya team has the w3.org site and take it from there, whether or not they are
adapted it for the Windows N T platform common in business, members of the consortium.
too. I now road test the latest versions of these tools as soon as I We create other tools as we need them, and our tool-creation
can get them, sending back crash reports on a bad day and occa- crew is always much in demand. Meeting registration, mailing-
sionally a bottle of champagne on a good one. list management, and control for our Web site are typical ex-
We are using our open source Java-based server, Jigsaw, fo r amples. We are looking forward to the time when we will use
collaborative work. For example, Jigsaw allows direct editing, Public key cryptography to authenticate collaborators. Every now
saves the various edited versions of a document, and keeps track a t l
d again the new systems go down, and we pay the price for
of what has been changed from one version to the next. I can call being on the bleeding edge by having to wait till they are fixed.

170 171
w e a v i n g t h e w e b m i n d t o m i n d

But we are gaining more of an understanding of what it will take The idea was that the state of the discussion would be visible to
1



to achieve the dream of collaboration through shared knowledge. everyone involved.
I would like any serious issue to be on the Web in hypertext.
I expect these tools to develop into a common new genre on the I would like annotation servers to exist where groups could add
Web. Real life is and must be full of all kinds of social con- links (or sticky yellow things) to documents they want to com-
straintâ€"the very processes from which "society" arises. Comput- ment on. Annotation servers are a third-party service allowing a
ers help if we use them to create abstract social machines on the group to share each others' coments on documents anywhere
Web: processes in which the people do the creative work and the else in the Web. The browser gets the original page and then sep-
machine does the administration. Many social processes can be arately checks annotation servers for comments, which are then
better run by machine, because the machine is always available, superimposed on the page. Imagine having servers for comments
it is free from bias, and no one likes to administer these kinds of in different forums, perhaps family, school, and company. Each
systems anyway. Online voting is one example, and it's already point and rebuttal is linked, so everyone can see at a glance the
beginning to happen: ADP Investor Communications and First direct agreements and contradictions and the supporting evi-
Chicago Trust have services that conduct online proxy voting for dence for each view, such that anything could be contested by
corporate shareholder meetings, and more than a thousand com- the people involved. If there was some sort of judicial, democra-
panies are using them. tic process for resolving issues, the discussion could be done in a
People are already experimenting with new social machines very clear and open fashion, with a computer keeping track of
for online peer review, while other tools such as chat rooms the arguments. Again, the theme is human beings doing the
developed quite independently and before the Web. M U D D s are thinking and machines helping it work on a larger scale, but
social tools derived from multiuser games of Dungeons and Drag- nothing replacing wisdom in the end.
ons where thousands of people take on roles and interact in a My hope was that the original "Discussion" idea, and future
global, online fantasy world. By experimenting with these struc- mechanisms that could evolve from it on the new Web, would
tures we may find a way to organize new social models that not move us beyond the historical situation of people hurling mud at
only scale well, but can be combined to form larger structures. each other, of peppering their arguments with personal abuse
Almost a decade ago now, I asked Ari Luotonen to spend and vitriol, and replace all that with much more of a reasoned,
three days writing a discussion tool for the nascent Web. It was to Socratic debate, in which individual ideas, accusations, and
be like a newsgroup, except that it would capture the logic of an Pieces of evidence can be questioned or supported.
argument. I'd always been frustrated that the essential role of a What Ari and I were trying to do was create a machine that
message in an argument was often lost information. When An would do the administration for, say, a court, or working group,
was done, anywhere on the C E R N server that we created a sub- ° Parliament. The initial trial was a discussion for the sake of
r



directory called Discussion, a new interactive forum would exist- 'scussion, and it didn't make a big splash. There are now a
It allowed people to post questions on a given subject, read, and nu
mber of software products for doing some of these things. To
respond. A person couldn't just "reply." He had to say whether he actually emulate a courtroom or a democratic voting process,
was agreeing, disagreeing, or asking for clarification of a point 0v
vever, the tools need much more development. I long for a

173
w e a v i n g t h e w e b
m i n d t o m i n d


move from argument by repetition of sound bites to a hypertext clean water and health careâ€"the necessity of those better off to
exposition that can be justified and challengedâ€"one that will care for but not simply control those less advantaged. I do no
allow us to look up and compare, side by side, what politicians more than touch on that urgent debate here.
or defendants and accusers, actually say, regardless of what is The stage-is set for an evolutionary growth of new social
claimed in television commercials and nightly news interviews. engines. The ability to create new forms of social process would
Because of low overhead, social machines will allow us to do be given to the world at large, and development would be rapid,
things we just couldn't do before. For example, they will allow us just as the openness of Web technology alldwed that to bloom.
to conduct a national plebiscite whose cost would otherwise be My colleagues and I have wondered whether we should seed
prohibitive. This would, of course, like all the benefits of this new this process using the consortium itself. We could construct the
technology, be biased toward those with Internet access. This is consortium social machine out of the many machines that make
just an example to show that we can reassess what is possible; I up working groups and staff meetings and so on. We could allow
am not advocating a move from representative democracy to a set of working groups that can be shown to form a tight self-
direct democracy. We should be careful not to do things just reliant cluster to secede and form a new peer "clone" consortium.
because they are possible. The rules would have to include more than a newsgroup-like
Perhaps the Web will enable more organic styles of manage- vote; budgets and contributions would have to balance, and
ment, in which groups within a company form in a local, rather responsibility would have to be accepted. In theory, we could
ad hoc fashion. They could be made self-forming like a news- then generalize this new social form. Then anyone could start a
group, but with constraints that ensure that whoever joins is consortium, when the conditions were right, by pushing a few
needed for the work of the company and is covered by sufficient buttons on the Web page of a virtual "consortium factory."
budget. Beyond that, the company doesn't have much conven-
tional structure/When someone has a task to perform, they asso-
ciate with whomever they need to get it done. People make
commitments and negotiate them between groups, without hav-
ing to go to a manager. The whole organic organization could
grow from a seed of a few digitally signed documents on the
Web, over the substrate of an electronic constitution that defines
how the social machines operate. Provisions for amending the
constitution would provide for mutation. A few minimalist rules
would ensure fairness.
While there is great excitement because these new social sys-
tems are essentially independent of geography, race, and religion
they will of course isolate those in developing countries who can-
not afford or have no option to access the Internet. At once the
great equalizer and the great divider, the Web highlightsâ€"as do

175
 C H A P T E R ; 3




M a c h i n e s a n d t h e W e b




Ln communicating between people using the Web, computers
and networks have as their job to enable the information space,
and otherwise get out of the way. But doesn't it make sense to
also bring computers more into the action, to put their analytical
power to work making sense of the vast content and human dis-
course on the Web? In part two of the dream, that is just what
they do.
The first step is putting data on the Web in a form that
machines can naturally understand, or converting it to that form.
This creates what I call a Semantic Webâ€"a web of data that can
he processed directly or indirectly by machines.
Consider the limited amount of help we have received so far
°n the Web from machines. Search engines have proven femark-
a
° l y useful in combing large indexes very rapidly and finding
obscure documents. But they have proven remarkably useless,
t o
° , in that they have no way to evaluate document quality. They

177
w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


return a lot of junk. The problem is that search engines generally current connections between the people, forums, and information
just look at occurrences of words in documentsâ€"something that sources. The structures and interrelations are important.
is a hint at but tells very little about what the document really i s The same sort of Web analysis could uncover new markets. It
or says. could help a project team leader evaluate the workings of her
A bit more sophisticated are automated brokerage services, team by mapping all the dependencies and relationships among
which began to emerge in 1998. These are Web sites that try to people, meeting minutes, research, and other materials involving
match buyers and sellers. From the buyer's perspective, such a the group, which together define how the project is going. A C E O
service can look like a metashopâ€"a store of stores. One metashop would like to be able to analyze his company's entire operation.
to emerge is webmarket.com: Give it a book title, and it will Imagine receiving a report along the lines of: "The company
search all the online bookstores it knows, check the prices, and looks fine, except for a couple of things. You've got a parts divi-
present a competitive list. To actually search the bookstores' cata- sion in Omaha that has exactly the same structure and business
logues, it has to pretend to be a browsing buyer, run then search patterns as a company in Detroit that just folded: You might want
engines, then extract the resulting data about product, price, and to look at that. There's a product you make that is completely
delivery. It can then prepare a table comparing each deal. documented but completely unused. And there seem to be a few
The trick of getting a computer to extract information from employees who are doing nothing that contributes .to the com-
an online catalogue is just that: a trick. It is known as screen pany at all."
scrapingâ€" trying to salvage something usable from information None of this analysis can be automated today, partly because
that is now in a form suitable only for humans. It is tenuous the form of intelligence that can draw such conclusions is diffi-
because the catalogue could change format overnightâ€"for ex- cult enough to find in people, yet alone in a computer program.
ample, putting the ISBN number where the price used to beâ€" But a simpler reason is that very little of the information on the
and the automatic broker would be confused. Web is in a form that a machine can understand. The Semantic
As people learn to use the Web, they analyze it in many ways. Web tackles this simpler problemâ€"perhaps in the end as a foun-
Ego surfingâ€"looking for occurrences of one's own nameâ€"is a dation for tackling the greater problem.
simple example. It may seem narcissistic, but it is a reasonable Today, when one person posts a notice on a Web site to sell,
quest, because we have a certain responsibility to figure out say, a yellow car, it is almost impossible for another person to
where we fit into the world. Online research is a more serious fmd it. Searching for a "yellow car for sale in Massachusetts"
example: One tries to find not only the answer to a question, but results in a useless huge list of pages that happen to contain those
also what structures might be out there in the information. words, when in fact the page I would want may be about a
Take a writer who wants to influence decision makers in Pak- Honda, good runner, any good offer" with a Boston phone num-
istan and India who are toying with the possible use of nuclear ber. The search engine doesn't understand the page, because it is
weapons. He wants to give them a deep awareness of the horrible written for a human reader with a knowledge of English and a lot
aftermath of the atomic bombing of Nagasaki. He needs to know °f common sense.
the forums in which these people operate, what they read. He This changes when the seller uses a program (or Web site)
needs sources of information on nuclear weapons. He needs the tQ
at allows him to fill out a form about an object for sale. This

178 179
w e a v i n g t-he w e b m a c h i n e s a n d t h e w e b


could result in a Web page, in a machine-readable format, that weather database. The relationships between the columns are the
maintains the significance of the document and its various parts. semantics- the meaning-of the data. These data are ripe for pub-
If all notifications of cars for sale were posted using the same lication as a semantic Web page. For this to happen, we need a
form, then it would be easy for search engines to find, exclu- common language that allows computers to represent and share

sively, yellow cars in Massachusetts. This is the simplest first step data, just as H T M L allows computers to represent and share

toward machine-understandable data. hypertext. The consortium is developing such a language, the

The next step is a search engine that can apply logic to Resource Description Framework (RDF), which, not surprisingly,
is based on X M L . In fact it is just X M L with some tips about
deduce whether each of the many responses it gets to an initial
which bits are data and how to find the meaning of the data. R D F
search is useful. This would allow us to ask general questions of
can be used in files on and off the Web. It can also be embedded
our computerized agents, such as "Did any baseball teams play
in regular H T M L Web pages. The R D F specification is relatively
yesterday in a place where the temperature was 22°C?" A pro-
basic, and is already a W 3 C Recommendation. What we need
gramâ€"call it a logic engineâ€"would apply mathematical reason-
now is a practical plan for deploying it.
ing to each item found. The search engine might find six
thousand facts involving baseball teams, and two million data The first form of semantic data on the Web was metadata-

items about temperatures and cities. The logic engine would ana- information about information. (There happens to be a company
called Metadata, but I use the term here as a generic noun, as it
lyze which bits of data refer to where a baseball team is, ascer-
has been used for many years.) Metadata consist of a set of prop-
tain what the temperature was in certain towns, filter both sets of
erties of a document. By definition, metadata are data, as well as
data, strip out all the junk, and respond: "The Red Sox played in
data about data. They describe catalogue information about who
Boston yesterday and the temperature was 22°C. Also, the Sharks
wrote Web pages and what they are about; information about
played in Tokyo, where it was 22°C." A simple search would have
how Web pages fit together and relate to each other as versions;
returned an endless list of possible answers that the human
translations, and reformattings; and social information such as
would have to wade through. By adding logic, we get back a cor-
distribution rights and privacy codes.
rect answer.
Most Web pages themselves carry a few bits of metadata.
While Web pages are not generally written for machines,
HTML pages have a hidden space in the document where certain
there is a vast amount of data in them, such as stock quotes and
"ems can be encoded, such as the page's title, its author, what
many parts of online catalogues, with well-defined semantics. I
software was used to create it, when it was created, and when it
take as evidence of the desperate need for the Semantic Web the
was last modified. Often this is also put in human-oriented form
many recent screen-scraping products, such as those used by the
| Plain English, at the bottom of a Web page in small type. Legal
n

brokers, to retrieve the normal Web pages and extract the origin 11


^formation, such as the copyright owner and privacy practice of
data. What a waste: Clearly there is a need to be able to go pub-
* e
Publisher, might be there, too. Metadata already out there
lish and read data directly.
so include catalogue information, such as keywords and classifi-
Most databases in daily use are relational databasesâ€" dat"
cation numbers, and all the things libraries tend to put on library
bases with columns of information that relate to each other, sue
c
ard . There is endorsement information, such as PICS labels.
s
as the temperature, barometric pressure, and location entries m
180 ltJl
 w e a v i n g t h e w e b m a c h i n e s a n d . t h e w e b


And there is structural information about which Web pages on 3 puter scientists at M I T and consortium members alike have
site act as cover page, table of contents, and index. There is n J °een known to raise their eyebrows and suggest that we should
end to metadata, and a common R D F language for metadata- ^ep the strength of the total language down. Should we, then,
should make a consistent world out of it. prevent the presence of powerfully descriptive languages on the
R D F ' s introduction has not been straightforward-and there Web?
has been a lot of discussion about how and even whether it The answer is that within many applications on the Web we
should be introduced. This is because, like many new languages should, but that in the Web as a whole we should not. Why?
it confronts a basic dilemma inherent in the design of any l a n ' Because when you look at the complexity of the world that the
guage. H T M L is a limiting language: You can use it only to Semantic Web must be able to describe, you realize that it must
express hypertext documents. Java, by contrast, isn't: You can b e possible to use any amount of power as needed. A reason for
write a bit of Java to do almost anything. Limiting languages are the success of the Web is that hypertext is so flexible a medium
useful because you can, for example, analyze an H T M L page ele- that the Web does not constrain the knowledge it tries to repre-
ment by element, convert it into other formats, index it, and sent. The same must be true for the web of meaning. I n fact, the
whatever. It is clear what every bit is for. People do all kinds of web of everything we know and use from day to day is complex:
things with H T M L pages that the pages were never originally We need the power of a strong language to represent it.
intended for. A Java applet is different. Because Java is a complete The trick here, though, is to make sure that each limited
programming language, you can use it to do anything, including mechanical part of the Web, each application, is within itself
creating a penguin that does somersaults. However, because Java composed of simple parts that will never get too powerful. In
is so powerful, the only way to figure out what a Java applet will many places we need the transparent simplicity of H T M L - s o
do is to run it and watch. When I designed H T M L for the Web, I each application, like an ATM machine, will work in a well-
chose to avoid giving it more power than it absolutely n e e d e d - a defined way. The mechanisms for metadata, privacy, payment,
"principle of least power," which I have stuck to ever since. I could and so on will all work in a well-defined way. The art of design-
have used a language like Donald Knuth's " T X , " which though it
e
ing applications in the future will be to fit them into the new
looks like a markup language is in fact a programming language. Web in all its complexity, yet make them individually simple
It would have allowed very fancy typography and all kinds of enough to work reliably every time. However, the total Web of all
gimmicks, but there would have been little chance of turning the data from each of the applications of R D F will make a very-
Web pages into anything else. It would allow you to express complex world, in which it will be possible to ask unanswerable
absolutely anything on the page, but would also have allowed questions. That is how the world is. The existence of such ques-
Web pages that could crash, or loop forever. This is the tension. tions will not stop the world from turning, or cause weird things
There is a fear that one day the big brother of R D F will to happen to traffic lights. But it will open the door to some very
become a programming language, and library cards will start interesting new applications that do roam over the whole
composing music, and checks will be made payable to a person intractable, incalculable Web and, while not promising anything,
whose name can be calculated only by using two hundred years deliver a lot.
of computer time. Looking at my plans for the Semantic Web,
182 183
w e a v i n g t h e w e b
m a c h i n e s a n d t h e w e b

To keep a given application simple, R D F documents can be lim¬
niinable discussions. Like words in the dictionary, everythingâ€"
ited so that they take on only certain forms. Every R D F docu-
until we tie things down to the physical worldâ€"is defined in
ment comes with a pointer at the top to its R D F schemaâ€" a
terms of other things.
master list of the data terms used in the document. Anyone can
This is also the basis of how computers can "understand"
create a new schema document. Two related schema languages
something. We learn very simple thingsâ€"such as to associate the
are in the works, one for X M L and one for RDF. Between them,
word hot with a burning feelingâ€"by early "programming" of our
they will tell any person or program about the elements of a Web
brains. Similarly, we can program a computer to do simple
page they describeâ€"for example, that a person's name is a string
things, like make a bank payment, and then we loosely say it
of characters but their age is a number. This provides everything
"understands" an electronic check. Alternatively, a computer
needed to define how databases are represented, and to start
could complete the process by following links on the Semantic
making all the existing data available. They also provide the tools
Web that tell it how to convert each term in a document it doesn't
for keeping the expressive power of an R D F document limited
understand into a term it does understand. I use the word seman-
and its behavior predictable. It allows us to unleash, bit by bit,
tic for this sort of machine-processible relative form of "mean-
the monster of an expressive language as we need it.
ing." The Semantic Web is the web of connections between
As the power is unleashed, computers on the Semantic Web different forms of data that allow a machine to do something it
achieve at first the ability to describe, then to infer, and then to wasn't able to do directly.
reason. The schema is a huge step, and one that will enable a
This may sound boring until it is scaled up to the entirety of
vast amount of interoperability and extra functionality. However,
the Web. Imagine what computers can understand when there is
it still only categorizes data. It says nothing about meaning or
a vast tangle of interconnected terms and data that can automati-
understanding.'
cally be followed. The power we will have at our fingertips will
People "come to a common understanding" by achieving a be awesome. Computers will "understand" in the sense that they
sufficiently similar set of consistent associations between words. will have achieved a dramatic increase in function by linking
This enables people to work together. Some understandings that very many meanings.
we regard as absolute truths, like the mathematical truth that a
To build understanding, we need to be able to link terms.
straight line is defined by two different points, are simple pat-
This will be made possible by inference languages, which work
terns. Other understandings, such as my understanding of some-
one level above the schema languages. Inference languages allow
one's anger at an injustice, are based on complex patterns of
computers to explain to each other that two terms that may seem
associations whose complete anatomy we are not fully aware of.
different are in some way the sameâ€"a little like an English-
When people "understand" something new, it means they can French dictionary. Inference languages will allow computers to
relate it to other things they already understand well enough. convert data from one format to another.
Two people from different planets can settle the difference
Databases are continually produced by different groups and
between red and blue by each making a prism, passing hgW
companies, without knowledge of each other. Rarely does anyone
through it, and seeing which color bends farther. But the differ- st
° p the process to try to define globally consistent terms for each
ence between love and respect will be hashed out only in inter- o f
the columns in the database tables. When we can link terms,
184
185
w e a v i n g t h e w e b


even many years later, a computer will be able to understand that' government can draw up its own tax form. As long as the infor-
what one company calls "mean-diurnal-temperature" is the same mation is in machine-understandable form, a Semantic Web pro-
as what another company calls "daily-average-temp." If HTML gram can follow semantic links to deduce that line 2 on the
and the Web made all the online documents look like one huge European form is like line 3A on the U.S. form, which is like line
book, RDF, schema, and inference languages will make all the 1 on the New York State tax form.
data in the world look like one huge database. Suppose I ask my computer to give me a business card for
When we have the inference layer, finding the yellow car for Piedro from Quadradynamics, but it doesn't have one. It can
sale becomes possible even if I ask for a yellow automobile. scan an invoice for his company name, address, and phone num-
When trying to fill in a tax form, my RDF-aware computer can ber, and take his e-mail address from a message, and present all
follow links out to the government's schema for it, find pointers the information needed for a business card. I might be the first
to the rules, and fill in all those lines for me by inference from to establish that mapping between fields, but now anyone who
other data it already knows. learns of those links can derive a business card from an e-mailed
invoice. If I publish the relationships, the links between fields, as
As with the current Web, decentralization is the underlying a bit of RDF, then the Semantic Web as a whole knows the
design principle that will give the Semantic Web its ability to equivalence.
become more than the sum of its parts. Forgive the simplified examples, but I hope the point is clear:
There have been many projects to store interlinked meanings Concepts become linked together. When, eventually, thousands
on a computer. The field has been called knowledge representation. of forms are linked together through the field for "family name"
These efforts typically use simple logical definitions such as the or "last name" or "surname," then anyone analyzing the. Web
following: A vehicle is a thing, a car is a vehicle, a wheel is thing, would realize that there is an important common concept here.
a car has four wheelsâ€"and so on. If enough definitions are The neat thing is that no one has to do that analysis. The concept
entered, a program could answer questions by following the links of "family name" simply begins to emerge as an important prop-
of the database and, in a mechanical way, pretend to think. The erty of a . per son. Like a child learning an idea from frequent
problem is that these systems are designed around a central data- encounters, the Semantic Web "learns" a concept from frequent
base, which has room for only one conceptual definition of "car." contributions from different independent sources. A compelling
They are not designed to link to other databases. note is that the Semantic Web does this without relying on Eng-
The Web, in contrast, does not try to define a whole system, lish or any natural language for understanding. It won't translate
just one Web page at any one time. Every page can link to every poetry, but it will translate invoices, catalogues, and the stuff of
other. In like fashion, the Semantic Web will allow different sites commerce, bureaucracy, travel, taxes, and so much more.
to have their own definition of "car." It can do this because the The reasoning behind this approach, then, is that there is no
inference layer will allow machines to link definitions. This central repository of information, and no one authority on any-
allows us to drop the requirement that two people have the same thing. By linking things together we can go a very long way
rigid idea of what something "is." In this way, the European Com- toward creating common understanding. The Semantic Web will
mission can draw up what it thinks of as a tax form. The U.S- Work when terms are generally agreed upon, when they are not,
186
187
 w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


and most often in the real-life fractal mess of terms that have 0 n it. Operating on such "partial understanding" is fundamental,
ious degrees of acceptance, whether in obscure helds or gl Q d W e do it all the time in the nonelectronic world. When
cultures. someone in Uruguay sends an American an invoice, the receiver
an't read most of it because it's in Spanish, but he can figure out
Making global standards is hard. The larger the number of peop| e that it is an invoice because it has references to a purchase-
who are involved, the worse it is. I n actuality, people can work order number, parts numbers, the amount that has to be paid,
together with only a few global understandings, and many lo ff ca a n d whom to pay. That's enough to decide that this is something
and regional ones. As with international and federal laws, and the he should pay, and to enable him to pay if. The two entities are
Web, the minimalist design principle applies: Try to constrain as operating wi,th overlapping vocabularies. The invoice is consis-
little as possible to meet the general goal. International com- tent with those drafted in Uruguay, and U . S . invoices are con-
merce works using global concepts of trading and debt, but it sistent as well, and there is enough commonality between
does not require everyone to use the same currency, or to have them to allow the transaction to be conducted. This happens
the same penalties for theft, and so on. with no central authority that mandates how an invoice must
Plenty of groups apart from W 3 C have found out how hard it be formulated.
is to get global agreement under pressure of local variations. As long as documents are created within the same logical
Libraries use a system called a M A R C record, which is a way of framework, such as RDF, partial understanding will be possible.
transmitting the contents of a library catalogue card. Electronic This is how computers will work across boundaries, without people
Data Interchange (EDI) was created a decade ago for conducting having to meet to agree on every specific term globally.
commerce electronically, with standard electronic equivalents of There will still be an incentive for standards to evolve,
things like order forms and invoices. In both cases, there was although they will be able to evolve steadily rather than by a
never complete agreement about all the fields. Some standards series of battles. Once an industry association, say, sets a stan-
were defined, but there were in practice regional or company- dard for metadata for invoices, business cards, purchase orders,
wide variations. Normal standards processes leave us with the shipping labels, and a handful of other e-commerce forms, then
impossible dilemma of whether we should have just one-to-one suddenly millions of people and companies with all sorts of com-
agreements, so that a Boeing invoice and an Airbus invoice are puters, software, and networks could conduct business electroni-
well defined but quite different, or whether we should postpone cally. Who will decide what the standard fields for an invoice
trying to do any electronic commerce until we define what an should be? Not the Web Consortium. They might arise in differ-
invoice is globally. ent ways, through ad hoc groups or individual companies or people.
The plan for the Semantic Web is to be able to move All the Web Consortium needs to do is set up the basic protocols
smoothly from one situation to another, and to work together that allow the inference rules to be defined, and each specialized
with a mixture. X M L namespaces will allow documents to work slice of life will establish the common agreements needed to
in a mixture of globally standard terms and locally agreed-upon make it work for them.
terms. The inference languages will allow computers to translate Perhaps the most important contribution of the Semantic Web
perhaps not all of a document, but enough of it to be able to act will be in providing a basis for the general Web's future evolution.
188 189
w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


The consortium's two original goals were to help the Web main- inference rules won't have to be a full programming language.
tain interoperability and to help it maintain "evolvability." \ty e They will be analyzable and separable, and should not present a
knew what we needed for interoperability. Evolvability was just a threat. However, for automating some real-life tasks, the language
buzzword. But if the consortium can now create an environment will have to become more powerful.
in which standardization processes become a property of how the Taking the tax form again, imagine that the instructions for
Web and society work together, then we will have created some- filling out your tax return are written in a computer language.
thing that not only is magic, but is capable of becoming ever The instructions are full of ifs and buts. They .include arithmetic,
more magical. and alternatives. A machine, to be able follow these instructions,
The Web has to be able to change slowly, one step at a time, will need a fairly general ability to reason. It will have to figure
without being stopped and redesigned from the ground up. This out what to put on each line by following links to find relation-
is true not only for the Web, but for Web applicationsâ€"the con- ships between data such as electronic bank statements, pay slips,
cepts, machines, and social systems that are built on top of it. and expense receipts.
For, even as the Web may change, the appliances using it will What is the advantage of this approach over, say, a tax-
change much more. Applications on the Web aren't suddenly cre- preparation program, or just giving in and writing a Java pro-
ated. They evolve from the smallest idea and grow stronger or gram to do it? The advantage of putting the rules in R D F is that
more complex. in doing so, all the reasoning is exposed, whereas a program is a
To make this buzzword concrete, just take that all too frequent black box: You don't see what happens inside it. When I used a
frustration that arises when a version-4 word processor comes tax program to figure out my 1997 taxes, it got the outcome
across a version-5 document and can't read it. The program typi- wrong. I think it got confused between estimated tax paid in
cally throws up its hands in horror at such an encounter with the 1997 and that paid for 1997, but I ' l l never know for sure. It read
future. It stops, because it figures (quite reasonably) that it cannot all my information and filled in the form incorrectly. I overrode
possibly understand a version-5 language, which had not been the result, but I couldn't fix the program because I couldn't see
invented when the program was written. However, with the infer- any of its workings. The only way I could have checked the pro-
ence languages, a version-5 document will be "self-describing." It gram would have been to do the job completely myself by hand.
will be provide a U R I for the version-5 schema. The version-4 pro- It a reasoning engine had pulled in all the data and figured the
gram can find the schema and, linked to it, rules for converting a taxes, I could have asked it why it did what it did, and corrected
version-5 document back into a version-4 document where possi- the source of the problem.
ble. The only requirement is that the version-4 software needs to Being able to ask "Why?" is important. It allows the user to
have been written so that it can understand the language in which trace back to the assumptions that were made, and the rules and
the rules are written. That R D F inference language, then, has to data used. Reasoning engines will allow us to manipulate, figure,
be a standard. hnd, and prove logical and numeric things over a wide-open field
When we unleash the power of R D F so that it allows us to °f applications. They will allow us to handle data that do not fall
express inference rules, we can still constrain it so that it is not 'nto clean categories such as "financial," "travel planning," and
such an expressive language that it will frighten people. The calendar." And they are essential to our trusting online results,

190 191
w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


because they will give us the power to know how the results \Veb. There have been projects in which a trust engine used a
were derived. less powerful language, but I honestly think that, looking at the
reality of life, we will need a very expressive language to express
The disadvantage of using reasoning engines is that, because they real trust, and trust engines capable of understanding such a lan-
can combine data from all over the Web in their search for an guage. The trick that will make the system work in practice will
answer, it can be too easy to ask an open question that will result be to send explanations around in most cases, instead of expect-
in an endless quest. Even though we have well-defined rules as to ing the receiver to figure out why it should believe something.
who can access the consortium's members-only Web site, one Creating the actual digital signature on" a document is the
can't just walk up to it and ask for admittance. This would ask simpler part of the trust technology. It can be done regardless of
the Web server to start an open-ended search for some good rea- the language used to create the document. It gives the ability to
son. We can't allow our Web server to waste time doing that; a sign a document, or part of a document, with a key, and to verify
user has to come equipped with some proof. Currently, users are that a document has been signed with a key. The plan is for a
asked under what rule, or through which member they have right standard way to sign any X M L document. The consortium in
of access. A human being checks the logic. We'd like to do it 1999 initiated this activity, combining earlier experience signing
automatically. In these cases we need a special form of R D F in PICS labels with new ideas from the banking industry.
which the explanation can be conveyedâ€"if you like, a statement The other part of trust, which actually weaves the Web of
with all the whys answered. While finding good argument for Trust, is the mesh of statements about who will trust statements
why someone should have access may involve large searches, or of what form when they are signed with what keys. This is where
inside knowledge, or complex reasoning, once that argument has the meat is, the real mirroring of society in technology. Getting
been found, checking it is a mechanical matter we could leave to¬ this right will enable everything from collaborating couples to
a simple tool. Hence the need for a language for carrying a proof commerce between corporations, and allow us actually to trust
across the Internet. A proof is just a list of sources for informa- machines to work on our behalf. As the Web is used to represent
tion, with pointers to the inference rules that were used to get more and more of what goes on in life, establishing trust gets
from one step to the next. more complicated. Right now, the real-life situation is too compli-
In the complexity of the real world, life can proceed even cated for our online tools.
when questions exist that reasoning engines can't answer. We just In most of our daily lives, then, even in a complex world,
don't make essential parts of our daily business depend on answer- each step should be straightforward. We won't have to unleash
ing them. We can support collaboration with a technical infrastruc- the full power of R D F to get our job done. There is no need to
ture that can respect society's needs in all their complexity. fear that using R D F will involve computers in guesswork.
Of course, our belief in each document will be based in the However, now that we are considering the most complex of
future on public key cryptography digital signatures. A "trust cases, we must not ignore those in which computers try to give
engine" will be a reasoning engine with a bolted-on signature reasonably good answers to open questions. The techniques they
checker giving it an inherent ability to validate a signature. The use are heuristicsâ€" ways of making decisions when all the alterna-
trust engine is the most powerful sort of agent on the Semantic tives can't be explored. When a person uses a search engine, and
192
193
 w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


casts her eye over the first page of returns for a promising l eac { people's eyes as I did in 1989, when I tried to explain how global
she is using a heuristic. Maybe she looks at the titles, or the fi rst
hypertext would work. But I've found a few individuals who
few lines quoted, or the UFJs themselves; in any case, usinp share the vision; I can see it from the way they gesticulate and
heuristics is an acquired art. Heuristic programs at a bank are the talk rapidly. I n these rare cases I also have that same gut feeling
ones that sound a warning when a person's credit-card spending a s I did a decade ago: They'll work for whomever they have to
pattern seems to differ from the usual. work for, do whatever it takes, to help make the dream come
The interplay between heuristic and strictly logical systems true. Once again, it's going to be a grassroots thing.'
will be interesting. Heuristics will make guesses, and logic will The blueprint for the new Web is also much like my 1989
check them. Robots will scan the Web and build indexes of certain proposal for the original Web. It has a social base, a technological
forms of data, and those indexes will become not definitive, but so plan, and some basic philosophy. A few people get it; most don't.
good that they can be used as definitive for many purposes. In the very beginning I wrote the World Wide Web code, then
Heuristics may become so good that they seem perfect. The went out into the world to promote the vision, made the technol-
Semantic Web is being carefully designed so that it does not have ogy freely available so people could start working on their little
to answer open questions. That is why it will work and grow. But piece of it, and encouraged them.
in the end it will also provide a foundation for those programs Today the consortium might write some of the code, or at
that can use heuristics to tackle the previously untacklable. least coordinate the writing of the code. Perhaps the computer
From here on it gets difficult to predict what will happen on community will share the vision and complete the pieces accord-
the Semantic Web. Because we will be able to define trust bound- ing to a business model that spans a number of years. Or perhaps
aries, we will be inclined, within those boundaries, to give tools someone watching from the sidelines will suddenly realize: "I
more power. Techniques like viruses and chain letters, which we know how I can do this. I don't know how to figure out a busi-
now think of as destructive, will become ways of getting a job ness model for it, but I think I can write the code in two weeks."
done. We will use heuristics and ask open questions only when Work on the first Web by people in various places progressed
; we have made a solid foundation of predictable ways of answer- in a fairly coordinated way because I had written the early code,
ing straightforward questions. We will be sorcerers in our new which gave other people something to write to. Now we have
world when we have learned to control our creations. two tools we didn't have then. One is the consortiumâ€"a place
where people can come together as well as a source of advanced
Even if the blueprint of technologies to achieve the new Web is software platforms like Jigsaw and Apache that people can use to
not crystal clear, the macroscopic view I've presented should at try out their new ideas. The second tool is the Web itself. Spread-
least convey that a lot of work has to be done. Some of it is far ing the word will be so much easier. I can publish this plan to the
along. Some of it is still a gleam in the eye. world even when it's only half finished. The normal academic
As work progresses, we will see more precisely how the way Robert Cailliaü and I could spread the original proposal was
pieces fit together. Right now the final architecture is hypotheti- to get it into the hypertext conference proceedingsâ€"and it was
cal; I'm saying it could fit together, it should fit together. When I rejected. This blueprint is not conference ready either, and I'm
try to explain the architecture now, I get the same distant look m not inclined to make it so. We'll just get the information out there

194 195
 w e a v i n g t h e w e b m a c h i n e s a n d t h e w e b


so people can point to it and discuss it. Once a seed is sown it going shopping i n a striped automobile on a Thursday. They pass
w i l l contain pointers back to where it came f r o m , so ideas w i l l the test of apparent novelty because there is no existing docu-
spread much more rapidly. ment describing exactly such a process. I n 1980, a device for
Cynics have already said to me, "You really think this time delivering a book electronically, or a device for online gambling,
around people are going to pick up on the architecture, and might have seemed novel, but now these things are just obvious
spend hours and hours on it as Pei Wei and all the others did?" Web versions of w e l l - k n o w n things. The U.S. Patent and Trade-
Yes. Because that's just what the cynics said i n 1989. They said, mark Office, ill-equipped to search for "prior art" (earlier occur-
"Oh, well, this is just too much to take on." But remember, it rence of the same idea) i n this new field, seems to have allowed
takes only a half dozen good people i n the right places. Back through patents by default.
then, finding that half dozen took a long time. Today, the w o r l d It is often difficult to know what a patent is about at all
can come to the consortium, plug i n their ideas, and have them because it is w r i t t e n obscurely using language quite different
disseminated. f r o m that w h i c h a normal programmer w o u l d use. There is a rea-
Indeed, the danger this time is that we get six hundred people son for this: The weapon is fear of a patent suit, rather than the
creating reasoning engines i n their garages across the land. But if patent itself. Companies cross-license patents to each other w i t h -
they t r y to patent what they're doing, each one of them thinking out ever settling i n court what those patents actually mean. Fear
they've found the grand solution first, or if they build palisades is increased by uncertainty and doubt, and so there is an incen-
of proprietary formats and use peculiar, undocumented ways of tive for obscurity. Only the courts can determine what a patent
doing things, they w i l l just get i n the way. If, through the consor- means, and the legal effort and time involved dwarfs the engi-
t i u m , hey come openly to the table for discussion, this could all neering effort.
w o r k out remarkably soon. This atmosphere is new. Software patents are new. The Inter-
net ethos i n the seventies and eighties was one of sharing for the
I mention patents i n passing, but i n fact they are a great stum- common good, and it w o u l d have been unthinkable for a player
bling block for Web development. Developers are stalling their to ask fees just for implementing a standard protocol such as
efforts i n a given direction w h e n they hear rumors that some HTTP. N o w things are changing. Large companies stockpile
company may have a patent that may involve the technology. patents as a threat of retaliation against suits f r o m their peers.
Currently, i n the U n i t e d States (unlike i n many countries), it is Small companies may be terrified to enter the business.
possible to patent part of the way a program does something. The lure of getting a cut of some fundamental part of the new
This is a little like patenting a business procedure: It is difficult to infrastructure is strong. Some companies (or even individuals)
define w h e n something really is "novel." Certainly among some make a living only by making up patents and suing larger compa-
patents I have looked at I have found it difficult to find anything nies, making themselves immune to retaliation by not actually
that gives me that "ah-ha" feeling of a new idea. Some just take a making or selling any products at all. The original aim of
well-known process (like interlibrary loan or betting on a race) patents â€"to promote the publication and deployment of ideas and
and do it i n software. Others combine w e l l - k n o w n techniques i n to protect the incentive for research â€"is noble, but abuse is now
apparently arbitrary ways to no added effect â€"like patenting a very serious problem.
196
197
 w e a v i n g t h e w e b


The ethos now seems to be that patents are a matter of what- C H A P T E R 14
ever you can get away w i t h . Engineers, asked by company
lawyers to provide patentable ideas every few months, resignedly
hand over "ideas" that make the engineers themselves cringe.
It is time for a change, to an ethos i n w h i c h companies use
patents to defend their o w n valid products, rather than serendipi-
tously suing based on claims even they themselves w o u l d have W e a v i n g t h e W e b
thought applied. The threshold of "innovation" is too low. Corpo-
rate lawyers are locked into a habit of arguing whatever advan-
tage they can, and probably only determined corporate leadership
can set the industry back on a sane track. The consortium mem-
bers have, at the time of w r i t i n g , been delivering on what to do,
but it is not clear what the result w i l l be.
The Semantic Web, like the Web already, w i l l make many
things previously impossible just obvious. As I w r i t e about the
new technology, I do wonder whether it w i l l be a technical
dream or a legal nightmare.


Can the future Web change the way people w o r k together and
advance knowledge i n a small company, a large organization, a
country? I f it works for a small group and can scale up, can it be
used to change the world? We know the Web lets us do things
more quickly, but can it make a phase change i n society, a move
to a new way of w o r k i n g â€" a n d w i l l that be for better or for
worse?
I n a company w i t h six employees, everybody can sit around a
table, share their visions of where they're going, and reach a
common understanding of all the terms they're using. I n a large
company, somebody defines the common terms and behavior that
make the company w o r k as an entity. Those w h o have been
through the transition know it only to w e l l : It typically kills
diversity. It's too rigid a structure. A n d it doesn't scale, because
as the company gets bigger, the bureaucratic boundaries cut off
more and more of its internal communications, its lifeblood. A t
198 199
w e a v i n g t h e w e b w e a v i n g t h e w e b


the other extreme is the U t o p i a n commune w i t h no structure, It's important that the Web help people be intuitive as well as
w h i c h doesn't w o r k either because nobody actually takes out the analytical, because our society needs both functions. H u m a n
garbage. beings have a natural balance i n using the creative and analytical
Whether a group can advance comes down to creating the parts of their brains. We w i l l solve large analytical problems by
right connectivity between people â€" i n a family, a company a turning computer power loose on the hard data of the Semantic
country, or the w o r l d . We've been trying to figure out how to cre- Web.
ate this for years. In many ways, we haven't had to decide, as Scaling intuition is difficult because our minds hold thou-
geography has decided for us. Companies, and nations, have sands of ephemeral tentative associations at the same time. To
always been defined by a physical grouping of people. The mili- allow group intuition, the Web w o u l d have to capture these

tary stability of a nation was based on troop placements and threads â€"half thoughts that arise, without evident rational

marching distances. The diversity of culture we've had also has thought or inference, as we w o r k . It w o u l d have to present them

stemmed f r o m two-dimensional space. The only reason t h e people to another reader as a natural complement to a half-formed idea.
The intuitive step occurs w h e n someone following links by a
i n a little village i n Switzerland w o u l d arise speaking a unique
number of independent people notices a relevant relationship,
dialect was that they were surrounded by mountains. Geography
and creates a shortcut link to record i t .
gave the w o r l d its military stability and cultural boxes. People
didn't have to decide how large their groups w o u l d be or where This all works only i f each person makes links as he or she

to draw the boundaries. Now that the metric is not physical dis- browses, so w r i t i n g , link creation, and browsing must be totally

tance, not even time zones, but clicks, we do have to make these integrated. I f someone discovers a relationship but doesn't make

decisions. The Internet and the Web have pulled us out of two- the link, he or she is wiser but the group is not.

dimensional space. They've also moved us away f r o m the idea To make such a shortcut, one person has to have t w o pieces

that we won't be interrupted by anybody who's more than a day's of inference i n his or her head at the same time. The new Web
w i l l make it much more likely that somebody somewhere is
march away.
browsing one source that has half of the key idea, and happens to
At first, this violation of our long-held rules can be unsettling,
have just recently browsed the other. For this to be likely, the
destroying a geographical sense of identity. The Web breaks the
Web must be w e l l connectedâ€"have few "degrees of separation."
boundaries we have relied on to define us and protect us, but it
This is the sort of thing researchers are always trying to do â€"get
can build new ones, too.
as m u c h i n their heads as possible, then go to sleep and hope to
The thing that does not scale w h e n a company grows is intu-
wake up i n the middle of the night w i t h a brilliant idea and rush
itionâ€"the ability to solve problems w i t h o u t using a well-defined
to write it down. But as the problems get bigger, we want to be
logical method. A person, or a small group brainstorming out
able to w o r k this brainstorming approach on a m u c h larger scale.
loud, ruminates about problems u n t i l possible solutions emerge.
We have to be sure to design the Web to allow feedback f r o m the
Answers arrive not necessarily f r o m following a logical p a t h , but
people who've made new intuitive links.
rather by seeing where connections may lead. A larger company
If we succeed, creativity w i l l arise across larger and more
fails to be intuitive w h e n the person w i t h the answer isn't talking
diverse groups. These high-level activities, w h i c h have occurred
w i t h the person w h o has the question.
200 201
w e a v i n g t h e w e b w e a v i n g t h e w e b


just within one human's brain, will occur among ever-larger individually, we have a choice in the social machines we create,
more interconnected groups of people acting as if they shared a t he variously shaped parts in our construction game. We know
larger intuitive brain. It is an intriguing analogy. Perhaps that that we want a well-connected structure for group intuition to
late-night surfing is not such a waste of time after all: It is just the •work. We know it should be decentralized, to be resilient and fair.
Web dreaming. The human brain outperforms computers by its incredible
level of parallel processing. Society, similarly, solves its problems
Atoms each have a valenceâ€"an ability to connect with just so in parallel. For the society to work efficiently on, the Web, mas-
many other atoms. As an individual, each of us picks a few chan- sive parallelism is required. Everybody must be- able to publish,
nels to be involved in, and we can cope with only so much. The and to control who has access to their published work. There
advantage of getting things done faster on the Web is an advan- should not be a structure (like a highway system or mandatory
tage only to the extent that we can accept the information faster, Dewey decimal system) or limitation that precludes any kind of
and there are definite limits. By just pushing the amount we have idea or solution purely because the Web won't allow it to be
to read and write, the number of e-mails we have to cope with, explained.
the number of Web sites we have to surf, we may scrape together The Internet before the Web thrived on a decentralized tech-
a few more bytes of knowledge, but exhaust ourselves in the nical architecture and a decentralized social architecture. These
process and miss the point. were incrementally created by the design of technical and social
As a group works together, the members begin to reach com- machinery. The community had just enough rules of behavior to
mon understandings that involve new concepts, which only they function using the simple social machines it invented. Starting
share. Sometimes these concepts can become so strong that the from a flat world in which every computer had just one Internet
group finds it has to battle the rest of the world to explain its address and everyone was considered equal, over time the sea of
decisions. At this point, the members may realize for the first chattering people imposed some order on itself. Newsgroups gave
time that they have started using words in special ways. They structure to information and people. The Web started with a simi-
may not realize how they have formed a little subculture until lar lack of preset structure, but soon all sorts of lists of "best"
they begin explaining their decisions to colleagues outside the sites created a competition-based structure even before advertis-
group. They have built a new understanding, and at the same ing was introduced. While the Internet itself seemed to represent
time built a barrier around themselves. Boundaries of under- a flight from hierarchy, without hierarchy there were too many
standing have been broken, but new ones have formed around degrees of separation to prevent things from being reinvented.
those who share the new concept. There seemed to be a quest for something that was not a tree, but
A choice has been made, and there is a gain and a loss in not a flat space, either.
terms of shared understanding. We certainly need a structure that will avoid those two cata-
What should guide us when we make these choices? What strophes: the global uniform McDonald's monoculture, and the
kind of a structure are we aiming for, and what principles will isolated Heaven's Gate cults that understand only themselves. By
help us achieve it? The Web as a medium is so flexible that it each of us spreading our attention evenly between groups of dif-
leaves the choice to us. As well as the choice of links we make ferent size, from personal to global, we help avoid these extremes.
202 203
w e a v i n g t h e w e b w e a v i n g t h e w e b


L i n k by link we build paths of understanding across the web of f r o m anything that a neuron could be aware of. From Arthur C.
humanity. We are the threads holding the world together. As we Clarke to Douglas Hofstader, writers have contemplated an "emer-
do this, we naturally end up w i t h a few Web sites i n very high gent property" arising from the mass of humanity and computers.
demand, and a continuum down to the huge number of Web sites But remember that such a phenomenon w o u l d have its o w n
w i t h only rare visitors. I n other words, appealing though equality agenda. We w o u l d not as individuals be aware of it, let alone con-
between peers seems, such a structure by its uniformity is not trol it, any more than the neuron controls the brain.
optimal. I t does not pay sufficient attention to global coordination, I expect that there w i l l be emergent properties w i t h the
and it can require too many clicks to get f r o m problem to solution. Semantic Web, but at a lesser level than emergent intelligence.
If instead everyone divides their time more or less evenly There could be spontaneous order or instability: Society could
between the top ten Web sites, the rest of the top one hundred, crash, much as the stock market crashed i n October 1987 because
the rest of the top one thousand, and so on, the load on various of automatic trading by computer. The agenda of tradingâ€"to
servers w o u l d have a distribution of sizes characteristic of "frac- make money on each trade â€"didn't change, but the dynamics did;
tal" patterns so common i n nature (from coastlines to ferns) and so many huge blocks of shares were traded so fast that the whole
of the famous "Mandelbrot set" mathematical patterns. I t turns system became unstable.
out that some measurements of all the Web traffic by Digital To ensure stability, any complex electronic system needs a
Equipment employees on the West Coast revealed very closely damping mechanism to introduce delay, to prevent it f r o m oscil-
this lln law: The Web exhibits fractal properties even though we lating too wildly. Damping mechanisms have since been built
can't individually see the patterns, and even though there is no into the stock-trading system. We may be able to build them into
hierarchical system to enforce such a distribution. the Semantic Web of cooperating computers â€"but w i l l we be able
This doesn't answer the question, but it is intriguing because to b u i l d them into the web of cooperating people? Already the
it suggests that there are large-scale dynamics operating to attention of people, the following of links, and the flow of money
produce such results. A fascinating result was found by Jon are interlaced inextricably.
Kleinberg, a computer scientist at Cornell University w h o dis- I do not, therefore, p i n m y hopes on an overpowering order
covered that, w h e n the matrix of the Web is analyzed like a emerging spontaneously f r o m the chaos. I feel that to deliberately
quantum mechanical system, stable energy states correspond to b u i l d a society, incrementally, using the best ideas w e have, is
concepts under discussion. The Web is starting to develop large- our duty and w i l l also be the most f u n . We are slowly learning
scale structure i n its o w n way. Maybe we w i l l be able to produce the value of decentralized, diverse systems, and of mutual respect
new metrics for checking the progress of society toward what and tolerance. Whether you put it down to evolution or your
we consider acceptable. favorite spirit, the neat thing is that we seem as humans to be
tuned so that we do i n the end get the most f u n out of doing the
The analogy of a global brain is tempting, because Web and brain "right" thing.
both involve huge numbers of elements â€"neurons and Web pages â€" M y hope and faith that we are headed somewhere stem i n
and a mixture of structure and apparent randomness. However, a part f r o m the repeatedly proven observation that people seem to
brain has an intelligence that emerges on quite a different level be naturally built to interact w i t h others as part of a greater

20/, 205
w e a v i n g t h e w e b w e a v i n g t h e w e b


system. A person who's completely turned inward, who spends low my weird reasoning to find out where it went wrong, but
all his or her time alone, is someone who has trouble making bal- would then use my own strange notation to explain the right
anced decisions and is very unhappy. Someone who is completely answer. This great feat involved looking at the world using my
turned outward, who's worried about the environment and inter- definitions, comparing them with his, and translating his knowl-
national diplomacy and spends no time sitting at home or in his edge and experience into my language. It was a mathematical
local community, also has trouble making balanced decisions and version of the art of listening. This sort of effort is needed when-
is also very unhappy. It seems a person's happiness depends on ever groups meet. It is also the hard work of the consortium's
having a balance of connections at different levels. We seem to working groups. Though it often seems to be: no fun, it is the
have built into us what it takes in a person to be part of a fractal thing that deserves the glory.
society. We have to be prepared to find that the "absolute" truth we
If we end up producing a structure in hyperspace that allows had been so comfortable with within one group is suddenly chal-
us to work together harmoniously, that would be a metamorpho- lenged when we meet another. Human communication scales up
sis. Though it would, I hope, happen incrementally, it would only if we can be tolerant of the differences while we work with.
result in a huge restructuring of society. A society that could partial understanding.
advance with intercreativity and group intuition rather than con- The new Web must allow me to learn by crossing boundaries.
flict as the basic mechanism would be a major change. It has to help me reorganize the links in my own brain so I can
If we lay the groundwork right and try novel ways of inter- understand those in another person's. It has to enable me to keep
acting on the new Web, we may find a whole new set of financial, the frameworks I already have, and relate them to new ones.
ethical, cultural, and governing structures to which we can Meanwhile, we as people will have to get used to viewing as
choose to belong, rather than having to pick the ones we happen communication rather than argument the discussions and chal-
to physically live in. Bit by bit those structures that work best lenges that are a necessary part of this process. When we fail, we
would become more important in the world, and democratic sys- will have to figure out whether one framework or another is bro-
tems might take on different shapes. ken, or whether we just aren't smart enough yet to relate them.
Working together is the business of finding shared under-
standings but being careful not to label them as absolute. They The parallels between technical design and social principles have
may be shared, but often arbitrary in the larger picture. recurred throughout the Web's history. About a year after I
We spend a lot of time trying to tie down meanings and fight- arrived to start the consortium, my wife and I came across Uni-
ing to have our own framework adopted by others. It is, after all, tarian Universalism. Walking into a Unitarian Universalist
a lifelong process to set ourselves up with connections to all the church more or less by chance felt like a breath of fresh air. Some
concepts we use. Having to work with someone else's definitions °f the association's basic philosophies very much match what I
is difficult. A n awe-inspiring talent of my physics tutor, Professor had been brought up to believe, and the objective I had in creat-
John Moffat, was that when I brought him a problem I h a
mg the Web. People now sometimes even ask whether I designed
worked out incorrectly, using a strange technique and symbols the Web based on these principles. Clearly, Unitarian Universal-
different from the well-established ones, he not only would K> l s
m had no influence on the Web. But I can see how it could

206 207
w e a v i n g t h e w e b w e a v i n g t h e w e b


have, because I did indeed design the Web around universalis! I was very lucky, in working at C E R N , to be in an environ-
(with a lowercase u) principles. ment that Unitarian Universalists and physicists would equally
'One of the things I like about Unitarianism is its lack of reli- appreciate: one of mutual respect, and of building something
gious trappings, miracles, and pomp and circumstance. It is mini- very great through collective effort that was well beyond the
malist, in a way. Unitarians accepted the useful parts of means of any one personâ€"without a huge bureaucratic regime.
philosophy from all religions, including Christianity and Judaism, The environment was complex and rich; any two people could
but also Hinduism, Buddhism, and any other good philosophies get together and exchange views, and even 'end up working
and wrapped them not into one consistent religion, but into an together somehow. This system produced a weird and wonderful
environment in which people think and discuss, argue, and machine, which needed care to maintain, but could take advan-
always try to be accepting of differences of opinion and ideas. tage of the ingenuity, inspiration, and intuition of individuals in a
I suppose many people would not classify "U-Uism" as a reli- special way. That, from the start, has been my goal for the World
gion at all, in that it doesn't have the dogma, and is very tolerant of Wide Web.
different forms of belief. It passes the Test of Independent Inven-
tion that I apply to technical designs: If someone else had invented Hope in life comes from the interconnections among all the
the same thing independently, the two systems should work people in the world. We believe that if we all work for what we
together without anyone having to decide which one was "central." think individually is good, then we as a whole will achieve more
For me, who enjoyed the acceptance and the diverse community of power, more understanding, more harmony as we continue the
the Internet, the Unitarian church was a great fit. Peer-to-peer rela- journey. We don't find the individual being subjugated by the
tionships are encouraged wherever they are appropriate, very- whole. We don't find the needs of the whole being subjugated by
much as the World Wide Web encourages a hypertext link to be the increasing power of an individual. But we might see more
made wherever it is appropriate. Both are philosophies that allow understanding in the struggles between these extremes. We don't
decentralized systems to develop, whether they are systems of expect the system to eventually become perfect. But we feel bet-
computers, knowledge, or people. The people who built the Inter- ter and better about it. We find the journey more and more excit-
net and Web have a real appreciation of the value of individuals ing, but we don't expect it to end.
and the value of systems in which individuals play their role, with Should we then feel that we are getting smarter and smarter,
both a firm sense of their own identity and a firm sense of some more and more in control of nature, as we evolve? Not really. Just
common good. better connectedâ€"connected into a better shape. The experience
There's a freedom about the Internet: As long as we accept the of seeing the Web take off by the grassroots effort of thousands
rules of sending packets around, we can send packets containing gives me tremendous hope that if we have the individual will, we
anything to anywhere. In Unitarian Universalism, if one accepts can collectively make of our world what we want.
the basic tenet of mutual respect in working together toward some
greater vision, then one finds a huge freedom in choosing one's
own words that capture that vision, one's own rituals to help focus
the mind, one's own metaphors for faith and hope.
208
209
P u b l i s h e r ' s N o t e : This appendix contains the original proposal for the World
Wide Web. At the author's request, it is presented here as a historical document
in its original state, with all of its original errors intactâ€"including typographical
and style elements â€"in order to preserve the integrity of the document.



I n f o r m a t i o n M a n a g e m e n t :

A P r o p o s a l

TIM BERNERS-LEE, CERN

MARCH 1 9 8 9, MAY 19 9 0



This proposal concents the management of general information about
accelerators and experiments at CERN. It discusses the problems of
loss of information about complex evolving systems and derives a solu-
tion based on a distributed hypertext system.




OVERVIEW
Many of the discussions of the future at CERN and the L H C
era end w i t h the question - "Yes, but how w i l l we ever keep track
of such a large project?" This proposal provides an answer to such
a p p e n d i x a p p e n d i x


questions. Firstly, it discusses the problem of information access at becomes available, and in order to get around unforeseen technical
C E R N . Then, it introduces the idea of linked information systems, problems. When a change is necessary, it normally affects only a
and compares them with less flexible ways of hnding information. small part of the organisation. A local reason arises for changing a
It then summarises my short experience with non-linear text sys- part of the experiment or detector. At this point, one has to dig

tems known as "hypertext", describes what C E R N needs from such around to find out what other parts and people will be affected.

a system, and what industry may provide. Finally, it suggests steps Keeping a book up to date becomes impractical, and the structure
of the book needs to be constantly revised.
we should take to involve ourselves with hypertext now, so that indi-
vidually and collectively we may understand what we are creating. The sort of information we are discussing answers, for exam-
ple, questions like
• Where is this module used?
LOSING INFORMATION AT CERN
C E R N is a wonderful organisation. It involves several thousand • Who wrote this code? Where does he work?
people, many of them very creative, all working toward common • What documents exist about that concept?
goals. Although they are nominaUy organised into a hierarchical man- • Which laboratories are included in that project?
agement structure,this does not constrain the way people will commu- • Which systems depend on this device?
nicate, and share information, equipment and software across groups. • What documents refer to this one?
The actual observed working structure of the organisation is a The problems of information loss may be particularly acute at
multiply connected "web" whose interconnections evolve with time. CERN, but in this case (as in certain others), C E R N is a model in
In this environment, a new person arriving, or someone taking on a miniature of the rest of world in a few years time. C E R N meets now
new task, is normally given a few hints as to who would be useful some problems which the rest of the world will have to face soon. In
people to talk to. Information about what facilities exist and how to 10 years, there may be many commercial solutions to the problems
ñnd out about them travels in the corridor gossip and occasional above, while today we need something to allow us to continue . 1




newsletters, and the details about what is required to be done spread
in a similar way. All things considered, the result is remarkably suc- LINKED INFORMATION SYSTEMS

cessful, despite occasional misunderstandings and duplicated effort. In providing a system for manipulating this sort of informa-
A problem, however, is the high turnover of people. When two tion, the hope would be to allow a pool of information to develop
years is a typical length of stay, information is constantly being lost. which could grow and evolve with the organisation and the pro-
The introduction of the new people demands a fair amount of their jects it describes. For this to be possible, the method of storage
time and that of others before they have any idea of what goes on. must not place its o w n restraints on the information.
The technical details of past projects are sometimes lost forever, or This is why a "web" of notes with links (like references)
only recovered after a detective investigation in an emergency. between them is far more useful than a fixed hierarchical system.
Often, the information has been recorded, it just cannot be found. When describing a complex system, many people resort to dia-
If a C E R N experiment were a static once-only development, all grams with circles and arrows. Circles and arrows leave one free to
the information could be written in a big book. As it is, C E R N is describe the interrelationships between things in a way that tables,
constantly changing as new ideas are produced, as new technology forexample, do not. The system we need is like a diagram of cir-
1
The same has been true, for example, of electronic mail gateways, document prepa-
212 "on, and heterogeneous distributed programming systems.
 a p p e n d i x
a p p e n d i x


cles and arrows, where circles and arrows can stand for anything. The system must allow any sort of information to be entered.
We can call the circles nodes, and the arrows links. Suppose Another person must be able to find the information, sometimes
each node is like a small note, summary article, or comment. I'm without knowing what he is looking for.
not over concerned here with whether it has text or graphics or In practice, it is useful for the system to be aware of the generic
both. Ideally, it represents or describes one particular person or types of the links between items (dependences, for example), and

object. Examples of nodes can be the types of nodes (people, things, documents..) without imposing

• People any limitations. ,

• Software modules
THE PROBLEM WITH TREES
• Groups of people
Many systems are organised hierarchically. The C E R N D O C
• Projects
documentation system is an example, as is the Unix file system,
• Concepts
and the VMS/HELP system. A tree has the practical advantage of
• Documents
• Types of hardware . giving every node a unique name. However, it does not allow the

• Specific hardware objects system to model the real world. For example, in a hierarchical
HELP system such as VMS/HELP, one often gets to a leaf on a tree
The arrows which links circle A to circle B can mean, for exam-
such as
ple, that A... H E L P C O M P I L E R SOURCE_FORMAT PRAGMAS DEFAULTS
• depends on B only to find a reference to another leaf: "Please see
• is part of B H E L P C O M P I L E R C O M M A N D OPTIONS DEFAULTS PRAG-
• made B MAS"
• refers to B and it is necessary to leave the system and re-enter it. What
• uses B was needed was a link from one node to another, because in this
• is an example of B
case the information was not naturally organised into a tree.
These circles and arrows, nodes and link 2, have different sig-
S
Another example of a tree-structured system is the uucp News
nificance in various sorts of conventional diagrams: system (try 'rn' under Unix). This is a hierarchical system of dis-
Nodes are Arrows mean
Diagram cussions ("newsgroups") each containing articles contributed by
People "Is parent of many people. It is a very useful method of pooling expertise, but
Family tree
Software modules "Passes data to" suffers from the inflexibility of a tree. Typically, a discussion under
Dataflow diagram
Module "Depends on" one newsgroup will develop into a different topic, at which point it
Dependency
Tasks "Must be done before"
PERT chart ought to be in a different part of the tree. (See Fig 1).
People "Reports to"
Organisational chart

from mcvax!uunet!pyrdc!pyrnj!rutgers!bellcore!geppetto!duncan T h u Mar...
Funked information systems hive entities and relattonsh.ps. There are h o w e ^ T m ^ ^«icle 93 of alt.hypertext:
differences between such a system and an 'Entity Relauonsiup" database s y s t e m ^
foth: cernvax!mcvax!uunet!pyrdc!pyrnj!rutgers!bellcore!geppetto!duncan
one thing the information stored in a linked system * largely comment for human fa
ers For another, nodes do not have strict types which define exactly what relauonsh.ps
215
they may have. Nodes of simialr type do not all have to be stored in the same place.
a p p e n d i x a p p e n d i x


> F r o m : duncan@geppetto.ctt.bellcore.com (Scott Duncan) hierarchical name. This particular note is expresses a problem with the
Newsgroups: alt.hypertext strict tree structure of the scheme: this discussion is related to several
Subject: Re: Threat to free information networks areas. Note that the "References", "From" and "Subject" fields can all be
Message-ID: < 14646@bellcore.bellcore.com > used to generate links.

Date: 10 M a r 89 21:00:44 G M T
References: < 1 7 8 4 . 2 4 1 6 B B 4 7 @ i s i s h q . F I D O N E T . O R G > THE PROBLEM WITH KEYWORDS

< 3437@uhccux.uhcc... Keywords are a common method of accessing data for which
Sender: news@bellcore.bellcore.com one does not have the exact coordinates. The usual problem with
Reply-To: duncan@ctt.bellcore.com (Scott Duncan) keywords, however, is that two people never chose the same key-
Organization: Computer Technology Transfer, Bellcore words. The keywords then become useful only to people who
already know the application well.
Lines: 18
Practical keyword systems (such as that of VAX/NOTES for
Doug T h o m p s o n has written what I felt w a s a thoughtful article on censorship example) require keywords to be registered. This is already a step
- m y acceptance or rejection of its points is not in the right direction.
particularly germane to this posting, however. A linked system takes this to the next logical step. Keywords
can be nodes which stand for a concept. A keyword node is then
I n reply G r e g Lee has somewhat tersely objected. no different from any other node. One can link documents, etc., to
keywords. One can then find keywords by finding any node to
M y question (and reason for this posting) is to ask w h e r e w e might logically which they are related. In this way, documents on similar topics
take this subject for more discussion. Somehow alt.hypertext does not seem
are indirectly linked, through their key concepts.

to be the proper place.
A keyword search then becomes a search starting from a small
number of named nodes, and finding nodes which are close to all
Would people feel it appropriate to move to alt.individualism or even one of of them.
the soc groups. I a m not so m u c h concerned w i t h the specific issue of censor- It was for these reasons that I first made a small linked infor-
ship of rec.humor.funny, but the v i e w s presented i n Greg's article. mation system, not realising that a term had already been coined
for the idea: "hypertext".

Speaking only for myself, of course, I a m . . .
Scott P. D u n c a n (duncan@ctt.bellcore.com O R ...!bellcore!ctt!duncan) A SOLUTION: HYPERTEXT

(Bellcore, 444 Hoes L a n e R R C 1H-210, Piscataway, NJ...)

(201-699-3910 (w) 201-463-3683 (h)| .
PERSONAL EXPERIENCE WITH HYPERTEXT
Fig 1. An article in the UUCP News scheme. In 1980, I wrote a program for keeping track of software with
which I was involved in the PS control system. Called Enquire, it
The Subject field allows notes on the same topic to be linked together flowed one to store snippets of information, and to link related
within a "newsgroup". The name of the newsgroup (alt.hypertext) is a P^ces together in any way. To find information, one progressed via

216 217
 a p p e n d i x
a p p e n d i x

the links from one sheet to another, rather like in the old computer Ã&lsqauo;nqui - I therefore produced a version for the V M S , and have
re




game "adventure". I used this for my personal record of people and us e d it to keep track of projects, people, groups, experiments, soft-

modules. It was similar to the application HyperCard produced ware modules and hardware devices with which I have worked. I

more recently by Apple for the Macintosh. A difference was that b a v e found it personally very useful. I have made no effort to

Enquire, although lacking the fancy graphics, ran on a multiuser m ake it suitable for general consumption, but have found that a

system, and allowed many people to access the same data. few people have successfully used it to browse through the pro-
jects and nnd out all sorts of things of their own accord.
Documentation o f t h e RPC p r o j e c t (concept)



Most of the documentation i s available on VMS, with the two HOT SPOTS
principle manuals being stored i n the CERNDOC system. Meanwhile, several programs have been made exploring these
ideas, both commercially and academically. Most of them use "hot
1) includes: The VAX/NOTES conference VXCERN::RPC spots" in documents, like icons, or highlighted phrases, as sensitive

2) includes: Test and Example suite
areas, touching a hot spot with a mouse brings up the relevant infor-
mation, or expands the text on the screen to include it. Imagine, then,
3) includes: RPC BUG LISTS
the references in this document, all being associated with the net-
4) includes: RPC System: Implementation Guide
work address of the thing to which they referred, so that while read-
Information for maintenance, porting, e t c .
ing this document you could skip to them with a click of the mouse.
5) includes: Suggested Development Strategy for RPC Applications
"Hypertext" is a term coined in the 1950s by Ted Nelson [...],
6) includes: "Notes on RPC", Draft 1, 20 feb 86
which has become popular for these systems, although it is used to
7) includes: "Notes on Proposed RPC Development" 18 Feb 86
embrace two different ideas. One idea (which is relevant to this
8) includes: RPC' User Manual
problem) is the concept:
How to build and run a distributed system.

9) includes: Draft Specifications and Implementation Notes 'Hypertext": Human-readable information linked together in an

10) includes: The RPC HELP f a c i l i t y unconstrained way
11) describes: THE REMOTE PROCEDURE CALL PROJECT i n DD/OC
The other idea, which is independent and largely a question of
Help Display Select Back Quit Mark Goto_mark Link Add Edit technology and time, is of multimedia documents which include
graphics, speech and video. I will not discuss this latter aspect fur-
Fig 2. A screen in an Enquire scheme.
ther here, although I will use the word "Hypermedia" to indicate
This example is basically a list, so the list of links is more that one is not bound to text.
important than the text on the node itself. Note that each link has It has been difficult to assess the effect of a large hypermedia
a type ("includes" for example) and may also have comment associ- system on an organisation, often because these systems never had
ated with it. (The bottom line is a menu bar.) seriously large-scale use. For this reason, we require large amounts
Soon after my re-arrival at C E R N in the D D division, I found of existing information should be accessible using any new infor-
that the environment was similar to that in PS, and I missed mation management system.
219
218
a p p e n d i x a p p e n d i x


CERN REQUIREMENTS DATA ANALYSIS
To be a practical system in the C E R N environment, there are a
An intriguing possibility, given a large hypertext database with
number of clear practical requirements. typed links, is that it allows some degree of automatic analysis. It is
possible to search, for example, for anomalies such as undocu-
REMOTE ACCESS ACROSS NETWORKS. mented software or divisions which contain no people. It is possi-
C E R N is distributed, and access from remote machines is ble to generate lists of people or devices for other purposes, such
essential. as mailing lists of people to be informed of changes. t



It is also possible to look at the topology of an organisation or a
HETEROGENEITY project, and draw conclusions about how it should be managed,
Access is required to the same data from different types of sys- and how it could evolve. This is particularly useful when the data-
tem (VM/CMS, Macintosh, VAX/VMS, Unix) base becomes very large, and groups of projects, for example, so
interwoven as to make it difficult to see the wood for the trees.
NON-CENTRALISATION In a complex place like CERN, it's not always obvious how to
Information systems start small and grow. They also start iso- divide people into groups. Imagine making a large three-dimen-
lated and then merge. A new system must allow existing systems sional model, with people represented by little spheres, and strings
to be linked together without requiring any central control or between people who have something in common at work.
coordination. Now imagine picking up the structure and shaking it, until you
make some sense of the tangle: perhaps, you see tightly knit
ACCESS TO EXISTING DATA groups in some places, and in some places weak areas of commu-
If we provide access to existing databases as though they were nication spanned by only a few people. Perhaps a linked informa-
in hypertext form, the system will get off the ground quicker. This tion system will allow us to see the real structure of the
is discussed further below. organisation in which we work.

PRIVATE LINKS LIVE LINKS
One must be able to add one's own private links to and from The data to which a link (or a hot spot) refers may be very sta-
public information. One must also be able to annotate links, as tic, or it may be temporary. In many cases at C E R N information
well as nodes, privately. about the state of systems is changing all the time. Hypertext
allows documents to be linked into "live" data so that every time
BELLS AND WHISTLES the link is followed, the information is retrieved. If one sacrifices
Storage of ASCII text, and display on 24x80 screens, is in the portability, it is possible so make following a link fire up a special
short term sufficient, and essential. Addition of graphics would be application, so that diagnostic programs, for example, could be
an optional extra with very much less penetration for the linked directly into the maintenance guide.
moment.


220 221
 a p p e n d i x a p p e n d i x


NON REQUIREMENTS THE "PERSONAL SKILLS INVENTORY".
Discussions on Hypertext have sometimes tackled the problem Personal skills and experience are just the sort of thing which
of copyright enforcement and data security. These are of secondary need hypertext flexibility. People can be linked to projects they
importance at C E R N , where information exchange is still more have worked on, which in turn can be linked to particular
important than secrecy. Authorisation and accounting systems for machines, programming languages, etc.
hypertext could conceivably be designed which are very sophisti-
cated, but they are not proposed here. THE STATE OF THE ART IN HYPERMEQIA

In cases where reference must be made to data which is in An increasing amount of work is being done into hypermedia
fact protected, existing file protection systems should be sufficient. research at universities and commercial research labs, and some
commercial systems have resulted. There have been two conferences,
SPECIFIC APPLICATIONS Hypertext '87 and '88, and in Washington D C , the National Institute
The following are three examples of specific places in which of Standards and Technology (NST) hosted a workshop on standardis-
the proposed system would be immediately useful. There are ation in hypertext, a followup of which will occur during 1990.
many others. The Communications of the ACM special issue on Hypertext con-
tains many references to hypertext papers. A bibliography on
DEVELOPMENT PROJECT DOCUMENTATION. hypertext is given in [NIST90], and a uucp newsgroup alt.hypertext
The Remote procedure Call project has a skeleton description exists. I do' not, therefore, give a list here.
using Enquire. Although limited, it is very useful for recording who
did what, where they are, what documents exist, etc. Also, one can BROWSING TECHNIQUES
keep track of users, and can easily append any extra little bits of Much of the academic research is into the human interface
information which come to hand and have nowhere else to be put. side of browsing through a complex information space. Problems
Cross-links to other projects, and to databases which contain infor- addressed are those of making navigation easy, and avoiding a feel-
mation on people and documents would be very useful, and save ing of being "lost in hyperspace". Whilst the results of the research
duplication of information. are interesting, many users at C E R N will be accessing the system
using primitive terminals, and so advanced window styles are not
DOCUMENT RETRIEVAL. so important for us now.
The C E R N D O C system provides the mechanics of storing and
printing documents. A linked system would allow one to browse INTERCONNECTION OR PUBLICATION?
through concepts, documents, systems and authors, also allowing Most systems available today use a single database. This is
references between documents to be stored. (Once a document accessed by many users by using a distributed file system. There
had been found, the existing machinery could be invoked to print are few products which take Ted Nelson's idea of a wide "docu-
it or display it). verse" literally by allowing links between nodes in different data-
bases. In order to do this, some standardisation would be
necessary. However, at the standardisation workshop, the empha-

222 223
 a p p e n d i x a p p e n d i x


sis was on standardisation of the format for exchangeable media This division also is important in order to allow the heterogeneity
nor for networking. This is prompted by the strong push toward which is required at C E R N (and would be a boon for the world in
publishing of hypermedia information, for example on optical disk
There seems to be a general consensus about the abstract data
model which a hypertext system should use.
Many systems have been put together with little or no regard
for portability, unfortunately. Some others, although published,
are proprietary software which is not for external release. How-
ever, there are several interesting projects and more are appear-
ing all the time. Digital's "Compound Document Architecture"
(CDA) , for example, is a data model which may be extendible
into a hypermedia model, and there are rumours that this is a
Information on

way Digital would like to go. one server reefers to
information on another


Fig 2. A client/server model for a distributed hypertext system.
INCENTIVES AND CALS
The US Department of Defence has given a big incentive to Therefore, an important phase i n the design of the system
hypermedia research by, in effect, specifying hypermedia docu- is to define this interface. After that, the development of various
mentation for future procurement. This means that all manuals forms of display program and of database server can proceed in par-
for parts for defence equipment must be provided in hypermedia allel. This will have been done well if many different information
form. The acronym CALS stands for "Computer-aided Acquisition sources, past, present and future, can be mapped onto the definition,
and Logistic Support). and if many different human interface programs can be written over
There is also much support from the publishing industry, and the years to take advantage of new technology and standards.
from librarians whose job it is to organise information.
ACCESSING EXISTING DATA
WHAT WILL THE SYSTEM LOOK LIKE? The system must achieve a critical usefulness early on.
Let us see what components a hypertext system at CERN must have. Existing hypertext systems have had to justify themselves solely on
new data. If, however, there was an existing base of data of person-
The only way in which sufficient flexibility can be incorporated nel, for example, to which new data could be linked, the value of
is to separate the information storage software from the information each new piece of data would be greater.
display software, with a well defined interface between them. What is required is a gateway program which will map an
Given the requirement for network access, it is natural to let this existing structure onto the hypertext model, and allow limited
clean interface coincide with the physical division between the user (perhaps read-only) access to it. This takes the form of a hypertext
and the remote database machine . 3
server written to provide existing information in a form matching
3 A client/server split at his level also makes multi-access more easy, in that a single the standard interface. One would not imagine the server actually
server process can service many clients, avoiding the problems of simultaneous access
225
to one database by many different users.
 a p p e n d i x a p p e n d i x



generating a hypertext database from and existing one: rather •
would generate a hypertext view of an existing database. Databases A generic tool could perhaps be made to allow any data-
base which uses a commercial DBMS to be displayed as a hyper-
text view.

In some cases, writing these servers would mean unscram-
bling or obtaining details of the existing protocols and/or file for-
mats. It may not be practical to provide the full functionality of
the original system through hypertext. In general, it will be more
important to allow read access to the general public: it may be
Fig 3. A hypertext gateway allows existing data to be seen in hypertext that there is a limited number of people who are providing the
form by a hypertext browser. information, and that they are content to use the existing facilities.
It is sometimes possible to enhance an existing storage system
Some examples of systems which could be connected in this by coding hypertext information in, if one knows that a server will
way are be generating a hypertext representation. In 'news' articles, for
example, one could use (in the text) a standard format for a refer-
uucp News This is a Unix electronic conferencing system. A server ence to another article. This would be picked out by the hypertext
for uucp news could makes links between notes on the same sub¬ gateway and used to generate a link to that note. This sort of
ject, as well as showing the structure of the conferences. enhancement will allow greater integration between old and new
VAX/Notes This is Digital's electronic conferencing system. It has a systems.
fairly wide following in FermiLab, but much less in C E R N . The There will always be a large number of information manage-
topology of a conference is quite restricting. ment systems - we get a lot of added usefulness from being able to
CERNDOC This is a document registration and distribution system cross-link them. However, we will lose out if we try to constrain
running on CERN's V M machine. As well as documents, categories them, as we will exclude systems and hamper the evolution of
and projects, keywords and authors lend themselves to representa- hypertext in general.
tion as hypertext nodes.
File systems This would allow any file to be linked to from other CONCLUSION
hypertext documents. We should work toward a universal linked information sys-
The Telephone Book Even this could even be viewed as hypertext, tem, in which generality and portability are more important than
with links between people and sections, sections and groups, peo- fancy graphics techniques and complex extra facilities.
ple and floors of buildings, etc. The aim would be to allow a place to be found for any infor-
The unix manual This is a large body of computer-readable text, cur- mation or reference which one felt was important, and a way of
rently organised in a flat way, but which also contains link infor- finding it afterwards. The result should be sufficiently attractive to
mation in a standard format ("See also.."). use that it the information contained would grow past a critical
226 227
 a p p e n d i x
a p p e n d i x

[SMISH88] Smish, J . B and Weiss, S.F/'An O v e r v i e w of Hypertext",in C o m -
threshold, so that the usefulness the scheme would in turn encour-
munications of the A C M , July 1988 Vol 31, No. 7,and other arti-
age its increased use.
cles i n the same special "Hypertext" issue.
The passing of this threshold accelerated by allowing large
[CAMP88] Campbell, B and Goodman, J , " H A M : a general purpose Hyper-
existing databases to be linked together and with new ones.
text Abstract Machine",in Communications of the A C M July 1988
Vol 31, No. 7
A PRACTICAL PROJECT
C




[ A S K C Y N 8 8 ] A k s c y n , R . M , M c C r a c k e n , D and Yoder E . A / ' K M S : A distributed
Here I suggest the practical steps to go to in order to find a real
hypermedia system for managing knowledge i n originations", i n
solution at C E R N . After a preliminary discussion of the require-
Communications of the A C M , July 1988 Vol 31, No. 7
ments listed above, a survey of what is available from industry is
[HYP88] Hypertext on Hypertext, a hypertext version of the special
obviously required. At this stage, we will be looking for a systems
C o m m s of the A C M edition, is avialble from the A C M for the
which are future-proof:
Macintosh or P C .
[RN] U n d e r unix, type m a n r n to find out about the rn c o m m a n d
• portable, or supported on many platforms,
w h i c h is used for reading uucp news.
• Extendible to new data formats.-
[NOTES] Under V M S , type HELP NOTES to find out about the
V A X / N O T E S system
We may find that with a little adaptation, pars of the system
[ C E R N D O C ] O n C E R N V M , type F I N D D O C F I N D for infrmation about h o w
we need can be combined from various sources: for example, a
to access the C E R N D O C programs.
browser from one source with a database from another.
[NIST90] J . Moline et. al. (ed.) Proceedings of the Hypertext Standardisa-
I : magine that two people for 6 to 12 months would be suffi-
tion Workshop January 16-18, 1990, National Institute of Stan-
cient for this phase of the project.
dards and Technology, pub. U . S . Dept. of C o m m e r c e
A second phase would almost certainly involve some program-
ming in order to set up a real system at C E R N on many machines.
An important part of this, discussed below, is the integration of a
hypertext system with existing data, so as to provide a universal
system, and to achieve critical usefulness at an early stage.
(... and yes, this would provide an excellent project with which
to try our new object oriented programming techniques!)


T B L March 1989, May 1990


REFERENCES
[NEL67] Nelson, T . H . "Getting it out of our system" i n Information
Retrieval: A Critical Review", G . Schechter, ed. Thomson Books,
Washington D . C . , 1967, 191-210
229
228
 G l o s s a r y




For background information and references for'this book, please
see http://www.w3.org/People/Berners-Lee/Weaving.


access control The ability to selectively control who can get at or
manipulate information in, for example, a Web server.
accessibility The art of ensuring that, to as large an extent as pos-
sible, facilities (such as, for example, Web access) are available
to people whether or not they have impairments of one sort or
another.
ACSS (Audio Cascading Style Sheets) A language for telling a
computer how to read a Web page aloud. This is now part of
CSS2.
Am ay a A n open source Web browser editor from W 3 C and
friends, used to push leading-edge ideas in Web client design.
Apache An open source Web server originally formed by taking
all the "patches" (hxes) to the NCSA Web server and making a
new server out of it.
browser A Web client that allows a human to read information on
the Web.
CERTM The European Particle Physics Laboratory, located on the
French-Swiss border near Geneva, Switzerland.
Click-stream Information collected about where a Web user has
been on the Web.
client Any program that uses the service of another program. On
the Web, a Web client is a program, such as a browser, editor,
or search robot, that reads or writes information on the Web.
CSS (Cascading Style Sheets) A W 3 C recommendation: a language
for writing style sheets. See also style sheet.
231
 g l o s s a r y g l o s s a r y


Cyc A knowledge-representation project in which a tree of defini- GIF (Graphics Interchange Format) A format for pictures transmit-
tions attempts to express real-world facts in a machine-readable ted pixel by pixel over the Net. Created by CompuServe, the
fashion. (Now a trademark of Cycorp Inc.) G I F specification was put into the public domain, but Unisys
digital signature A very large number created in such a way that found that it had a patent on the compression technology used.
it can be shown to have been done only by somebody in pos- This stimulated the development of PNG.
session of a secret key and only by processing a document with G1LC (Global Internet Liberty Campaign) A group that has been
a particular content. It can be used for the same purposes as a laudably vocal in support of individual rights on the Net (though
person's handwritten signature on a physical document. Some- occasionally tending to throw out the baby with:the bathwater),
thing you can do with public key cryptography. W 3 C work graphics Two- or three-dimensional images, typically drawings or
addresses the digital signature of X M L documents. photographs. See also GIF, PNG, SVG, and V R M L .
DOM (Document Object Model) Within a computer, information is HTML (Hypertext Markup Language) A computer language for rep-
often organized as a set of "objects." When transmitted, it is sent as resenting the contents of a page of hypertext; the language that
a "document." The D O M is a W3C specification that gives a com- most Web pages are currently written in.
mon way for programs to access a document as a set of objects, HTTP (Hypertext Transfer Protocol) A computer protocol for trans-
domain name A name (such as "w3.org") of a service, Web site, or ferring information across the Net in such a way as to meet
computer, and so on in a hierarchical system of delegated the demands of a global hypertext system. Part of the original
authorityâ€"the Domain Name System. design of the Web, continued in a W 3 C activity, and now a
DTD In the SGML world, a DTD is a metadocument containing HTTP 1.1 I E T F draft standard,
information about how a given set of SGML tags can be used. In hypertext Nonsequential writing; Ted Nelson's term for a
the X M L world this role will be taken over by a schema. Some- medium that includes links. Nowadays it includes other media
times, but arguably, "document type definition." See also schema. apart from text and is sometimes called hypermedia,
Dublin Core A set of basic metadata properties (such as title, etc.) information space The abstract concept of everything accessible
for classifying Web resources. using networks: the Web.
EBT (Electronic Book Technology) A company started by Andries 1 N K 1 A (lnstitut National de Recherche en lnfomatique et Automa¬
Van Dam and others to develop hypertext systems. tique) The French national research laboratory for computer
EDI (Electronic Data Interchange) A pre-Web standard for the elec- science and control. Cohost of W 3 C and developers of Amaya.
tronic exchange of commercial documents. Internet A global network of networks through which computers
Enquire A 1980 program, named after the Victorian book Enquire communicate by sending information in packets. Each network
Within upon Everything. consists of computers connected by cables or wireless links.
filtering The setting up of criteria to select a subset of data from Intranet A part of the Internet or part of the Web used internally
a broad stream of it. Filtering information is essential for every- within a company or organization.
one in daily life. Filtering by parents of small children may be IP (Internet Protocol) The protocol that governs how computers
wise. Filtering by o t h e r s - I S P s or governments-is bad, and is send packets across the Internet. Designed by Vint Cerf and
called censorship. Bob Khan. (IP may also stand for intellectual property; see IPR.)
232 233
 g l o s s a r y g l o s s a r y


1PR (Intellectual Property Rights) The conditions under which the sion with the computer program. If you have seen a "DOS win-
information created by one party may be appreciated by dow," then you have some idea of how people did their commu-
another party. nicating with computers in those days, before they learned how
ISO (International Standards Organization) A n international group to drag and drop. Line-mode is still a very respectable way to
of national standards bodies. communicate with a computer.
ISP (Internet service provider) The party providing one with con- line-mode browser A Web client that communicated with the user
nectivity to the Internet. Some users have a cable or some sort in line-mode and could run all kinds of computers that did not
of wireless link to their ISP. For others, their computer may dial have windows or mice.
an ISP by phone and send and receive Internet packets over the link A reference from one document to another (external link), or
phone line; the ISP then forwards the packets over the Internet. from one location in the same document to another (internal
Java A programming language developed (originally as "Oak") by link), that can be followed efficiently using a computer. The
James Gosling of Sun Microsystems. Designed for portability unit of connection in hypertext.
and usability embedded in small devices, Java took off as a lan- MARC record A standard for machine-readable library catalogue
guage for small applications ("applets") that ran within a Web cards.
browser. meta- A prefix to indicate something applied to itself; for exam-
Jigsaw Open source Web server of great modularity, written in ple, a metameeting is a meeting about meetings.
Java. From W 3 C and friends. metadata Data about data on the Web, including but not limited
JPEG (Joint Photographic Experts Group) This group defined a for- to authorship, classification, endorsement, policy, distribution
mat for encoding photographs that uses fewer bytes than the terms, IPR, and so on. A significant use for the Semantic Web.
pixel-by-pixel approaches of G I F and PNG, without too much micropayments Technology allowing one to pay for Web site
visible degradation in quality. The format (JFIF) is casually • access in very small amounts as one browses.
referred to as J P E G . minimal constraint, principle of The idea that engineering or other
Keio University Near Tokyo, Japan. Cohost of W3C. designs should define only what they have to, leaving other
LCS (Laboratory for Computer Science) A laboratory at the Massa- aspects of the system and other systems as unconstrained as
chusetts Institute of Technology. Cohost of W 3 C . possible.
LEAD (Live Early Adoption and Demonstration) A W3C policy to eat MIT (Massachusetts Institute of Technology) See L C S . Cohost of
our own cooking to find out how it can be better. W3C.
libwww The library (collection) of WWW-related program mod- mobile devices Pagers, phones, handheld computers, and so on.
ules available for free use by anyone since the start of the Web. All are potentially mobile Internet devices and Web clients.
line-mode In high and far-off times, people did not see computer Mosaic A Web browser developed by Marc Andreessen, Eric
programs through windows. They typed commands on a termi- Bina, and their colleagues at NCSA.
nal, and the computer replied with text, which was displayed NCSA (National Center for Supercomputing Applications) A center at
on the screen (or printed on a roll of paper) interleaved with the University of Illinois at Urbana-Champaign whose software
the commands, much as though the person were in a chat ses- development group created Mosaic.
234 235
 g l o s s a r y g l o s s a r y



Nelson, Ted Coiner of the word hypertext; guru and visionary. PK1 (Public Key Infrastructure) A hierarchy of "certification authori-
Net Short for Internet. ties" to allow individuals and organizations to identify each other
NeXT Name of the company started by Steve Jobs, and of the for the purpose (principally) of doing business electronically.
computer it manufactured, that integrated many novelties such PNG (Portable Network Graphics) A format for encoding a picture
as the Mach kernel, Unix, NeXTStep, Objective-C, drag-and- pixel by pixel and sending it over the Net. A recommendation
drop application builders, optical disks, and digital signal of the W3C, replacing GIF.
processors. The development platform I used for the hrst Web protocol A language and a set of rules that alkrw computers to inter-
client. act in a well-defined way. Examples are FTP, HTTP, and NNTP.
NNTP (Network News Transfer Protocol) A protocol that defines RDF (Resource, Description Framework) A framework for construct-
how news articles are passed around between computers. Each ing logical languages that can work together in the Semantic
computer passes an article to any of its neighbors that have not Web. A way of using X M L for data rather than just
yet got it. documents.
node Thing joined by links. In the Web, a node is a Web page, RPC (remote procedure call) When one part of a program calls on
any resource with a U R I . another part to do some work, the action is called a procedure
open source Software whose source code is freely distributed and call. R P C is a set of tools that allow you to write a program
modifiable by anyone. W3C sample code is open source soft- whose different parts are on different computers, without hav-
ware. A trademark of opensource.org. ing to worry about how the communication happens. A generic
packet A unit into which information is divided for transmission technique, not a specific product.
across the Internet. RSA A public key encryption system invented by Ron Rivest, Adi
partial understanding The ability to understand part of the import Shamir, and Leonard Adleman. RSA algorithms have been
of a document that uses multiple vocabularies, some but not all patented, and so its inventors have licensed its deployment.
of which are understood. schema (pi., schemata) A document that describes an X M L or
PGP (Pretty Good Privacy) An e-mail security system that uses R D F vocabulary.
public key cryptography and has the philosophy that individu- Semantic Web The Web of data with meaning in the sense that a
als can choose whom they trust for what purposeâ€"the "web of computer program can learn enough about what the data
trust." means to process it.
PICS (Platform for Internet Content Selection) W3C's technology separation of form from content The principle that one should
that allows parents to select content for their children on the represent separately the essence of a document and the style
basis of an open set of criteria, as opposed to government cen- with which it is presented. An element in my decision to use
sorship. See filtering. SGML and an important element in the drive for accessibility
PKC (public key cryptography) A very neat bit of mathematics on on the Web.
which is based a security system in which there is no need to server A program that provides a service (typically information)
exchange secret keys; instead, people have one "private" key to another program, called the client. A Web server holds Web
that only they know and one "public" key that everyone knows. pages and allows client programs to read and write them.
236 237
 g l o s s a r y g l o s s a r y


SGML (Standard Generalized Markup Language) An international tual Reality Markup Language," and implemented by Mark
standard in markup languages, a basis for H T M L and a precur- Pesce as a variant of Silicon Graphics's "Inventor" format; later
sor to X M L .
managed by the V R M L consortium, now "Web 3D" consortium.
SMIL (Synchronized Multimedia Integration Language) A language W3C (World Wide Web Consortium) A neutral meeting of those to
for creating a multimedia presentation by specifying the spatial whom the Web is important, with the mission of leading the
and temporal relationships between its components. A W3C
Web to its full potential.
recommendation.
WA1 (Web Accessibility Initiative) A domain of W 3 C that attempts
style sheet A document that describes to a computer program to ensure the use of the Web by anyone regardless of disability.
(such as a browser) how to translate the document markup into
WA1S (Wide Area Information Servers) A distributed information
a particular presentation (fonts, colors, spacing, etc.) on the
system designed by Brewster Kahle while at Thinking
screen or in print. See also CSS, X S L , separation of form from
Machines. WAIS was like a Web of search engines, but without
content.
hypertext.
SVG (Scalable Vector Graphics) A language for describing drawings Web Short for World Wide Web.
in terms of the shapes that compose them, so that these can be
World Wide Web (three words; also known as WWW) The set of all
rendered as well as possible.
information accessible using computers and networking, each
Tangle A program I wrote for playing with the concept of infor- unit of information identified by a U R I .
mation as consisting only of the connections.
WorldWideWeb (one word; no spaces) The name of the first Web
TCP (Transmission Control Protocol) A computer protocol that
client, a browser/editor that ran on a N e X T machine.
allows one computer to.send the other a continuous stream of
X The X Window system, invented by Bob Scheifler; a standard
information by breaking it into packets and reassembling it at
interface between a program and a screen that was ubiquitous
the other end, resending any packets that get lost in the Inter-
on Unix systems. Unlike Microsoft's Windows, from the begin-
net. T C P uses IP to send the packets, and the two together
ning X allowed programs running on one machine to display on
are referred to as TCP/IP.
another, across the Internet. Scheifler ran the X Consortium
URI (Universal Resource Identifier) The string (often starting with
from MIT/LCS for many years, then spun it off, and eventually
http:) that is used to identify anything on the Web.
closed it.
URL (Uniform Resource Locator) A term used sometimes for cer-
Xanadu Ted Nelson's planned global hypertext project.
tain URIs to indicate that they might change.
X M L (Extensible Markup Language) A simplified successor to
Viola An interpreted computer language (like Java) developed by
S G M L . W3C's generic language for creating new markup lan-
Pei Wei at the University of Berkeley. Also, a Web browser built
guages. Markup languages (such as H T M L ) are used to repre-
using Viola.
sent documents with a nested, treelike structure. X M L is a
virtual hypertext Hypertext that is generated from its U R I by a
product of W 3 C and a trademark of MIT.
program, rather than by recourse to a stored file.
X S L (Extensible Style Sheet Language) A style sheet language, like
VRML (Virtual Reality Modeling Language) A n idea for 3D compo-
CSS, but also allowing document transformation.
sitional graphics on the Web, proposed by Dave Raggett as "Vir-
238
239
 I n d e x




Abramatic, Jean-François, 102, 109 WorldWideWeb program released to
Addis, Louise, 45-46, 63 C E R N people, 45, 169; WorldWideWeb
Addressing scheme, 20. See also URL (URI) program released on Internet, 46
Adleman, Leonard, 149 Better Business Bureau Online, 148
America Online (AOL], 92, 105-6, 112, Bina, Eric, 68, 70-71, 77
124, 132 Bosak, Jon, 119
Andreessen, Marc,.68-71, 77, 82, 83, 93, Bray, Tim, 119
96,99-100,116,165 Brown, Peter, 26
Apple computer: HyperCard, 20; Macin- Browser. See Web browser
tosh browsers designed, 58, 67, 71, 77, Bruce, Tom, 69-70, 76, 77
99 Bush, Vannevar, 5, 6
Applets, 56, 104
"As We May Think" (Bush], 5 C language, 32; versus objective-C, 48
Cailliau, Robert, 25-26, 29, 31, 42, 44,
Baird-Smith, Anselm, 120 45, 50-51, 55, 56, 58, 74, 83, 88, 101,
Bangemann, Martin, 88, 95 157, 195; first international WWW con-
Barksdale, Jim, 103, 106 ference, 79-80, 85-89, 124; Macintosh
Barran, Paul, 6 browser, 58; president of WWW Con-
Berners-Lee, Tim: browser for WWW ference Committee, 92
created, 30; C E R N 1980, 4, 8-11; Cascading Style Sheets (CSS), 168
C E R N 1984-89, 12, 13-23; C E R N Case, Steve, 105-6, 112
1990-92, 31-32, 42-43, 45/46, 209; Censorship, 134-36; bias, 125; filters,
C E R N proposal (birth of Web), 21-22; 125, 134-35, 136; versus self-regula-
children born, 30-31, 87; code for tion, 113-14, 124. See also PICS; Filter-
WWW written, 28-29; director of ing software
W3C, 95, see also World Wide Web Cerf, Vint, 6
Consortium; earliest ideas for WWW, 1, C E R N , 4, 4n, 7-9, 14, 19; Berners-Lee
3-6, 23; education, 4, 84; Enquire, 1, 4, at, 4, 8-11, 12, 13-23, 31-32, 42-43,
9-11, 15-16, 17, 19; evangelizing for 46, 48, 60, 82; computers and compet-
the Web, 31, 42-43, 44, 50, 55; feelings ing software used at, 8-9, 14, 15, 22,
about commercial opportunities and 32, 43; Discussion subdirectory,
• WWW, 83-85, 107-8; first employ- 172-73; hosts international WWW con-
ment, 4; hypertext protocols, 140-41; ference, 79-81, 85-89; info.cern.ch
Internet enters life, 19, 22; at MIT's (server), 29, 32, 44, 45, 49, 55, 66, 75,
LCS, 72, 87-89, 91-102; NeXT PC, 95, 98; Internet and, 18-19; intranet,
22-23, 28, 31, 46, 49, 55; parents, 3-4; 57; Large Hadron Collider, 101; Large
sabbatical, 1992, 60-66; Tangle, 12-13; Electron Positron accelerator, 14; Norsk
television interview, 114-15; Unitarian Data SYNTAN-III operating system at,
Universalism and, 207-9; vision for 11; public domain status of WWW, 73;
WWW, 1-2, 27, 31, 33, 76, 83-84, 87, telephone book, 32; WWW's birth at,
91, 123, 157-75, 199-209; WWW 32-33; WWW Consortium and, 88, 97,
consortium, 75-76, 78, 81-89, 91-102; 101; WWW within, 46, 55

2/.1
 i n d e x
i n d e x


Clark, Dave, 54 IBM, 94, 112; e-business mark, 137-38 Lie, Hâkon, 116, 168
Form, separation of content from, 130¬
Clark, Jim, 82, 83, 93, 103 Image Computer Systems, 11 Literary Machines (Nelson), 5, 65
32, 168
C O M D E X , 1994, 93 Incompatiblity between computers and Links: analogy to brain, 37-38, 204-5;
Free Software Foundation, 73
Communications Decency Act, 113-14 Web, 35; Java and, 104 embedded, 139-40; in Enquire, 10,
FTP (file transfer protocol), 30, 38, 39
Compaq, 103, 131 infodesign.ch, 49 15-16; external, 10-11, 16, 33-34; free
59, 76
Information Sciences Institute, 40 speech and, 139-41; hot buttons, 20;
CompuServe, 92, 105, 132, 133-34, 165
INRIA (Institut National de Recherche en idea for, 3-4, 5, 12; information system
Connolly, Dan, 111, 119 Gates, Bill, 92-93, 108, 112
Informatique et en Automatique); Grif at, and, 21; internal, 10, 33; intuition and,
Content, separation of form from, Gateways, 49-50
44, 101, 119; W3C European base, 101-2 201; myths about, 140-41; normal,
130-32, 168 General Public License (GPL|, 73
International Computers Ltd., 72 139; paper, traditional, 38; program for,
Cyber Patrol, 135 G I F (Graphic Interchange Format),
International Standards Organization 19-20 -
165-66
Databases, 180-81, 185-86 (ISO), 17 Location independence, 159, 160
Gifford, David, 72
Davis, Donald, 6 Internet, 6, 17-18; access, changes Long, Dave, 81
Global Internet Liberty Campaign
D E C (Digital Equipment Corporation): needed in, 158-59; in America, 1980s, Ludvigsen, Borre, 86
(GILC), 136
C E R N visit, 77-78, 109; Consortium 17; -based information system, 40, 162; Luotonen, Ari, 58 - 59, 97, 172 - 73
Gopher, 40, 67, 72, 76, 84
and, 93-84; DECnet, 19, 20; intranet, Gosling, James, 104 Berners-Lee and, 19; decentralized
57 sociotechnical architecture, 203; Euro- MARC record, 188
Grant, Gail, 78
Decentralization, 16, 36, 99, 186, 203 pean alternative, 1980s, 17; hypertext Massinter, Larry, 54-55
Graphics formats, 1,65-67, 168
Dertouzos, Michael, 72, 76, 81-82, 91, and, 23, 26, 44-45; protocols and, 18, Mbeki, Thabo, 102
Grif, 44, 101, 102, 119
102 19, 38, 40; security issues, 97; service McCahill, Mark, 40
Groff, Jean-François, 48, 49
D. G . Nash Ltd., 4 providers (ISPs), 80-81, 109, 131, 133; Merit Inc., 80
Gross, Phil, 84
Documentation systems, 15-16, 19 use analyzed, 204; WorldWideWeb pro- Metadata, 181-82
Domain name system, 126-29; security gram released on, 46 Metakides, George, 88
Hagino, Tatsuya, 116
of, 150 Internet Assigned Numbers Authority Microsoft, 92-93; antitrust suit,
Hardin, Joseph, 70, 79-80
(I ANA), 127 116-17, 124, 130; Internet Explorer,
Dougherty, Dale, 64, 76-77, 96, 109, 119 Harman, Amy, 136
Internet Engineering Task Force (IETF), 108, 116; Netscape and, 93, 103; Win-
Dynatext, 27 Helsinki University of Technology, 56
53-54, 71, 73, 84, .92; meeting 1992; dows, 99; Windows 95 and bundled
Heuristics, 193-94
61-63 Internet access, 93, 108; Windows 98,
Ego surfing, 178 Hewlett-Packard (HP), 67-68, 94
116-17
Electronic Book Technology, 27 Hoesl, van, Frans, 59 Intranet, 57, 162 - 63
Minimal constraint, principle of, 39, 124
Electronic commerce (e-commerce), 2, Hypertext: alt.hypertext, WorldWideWeb
Java, 56, 104-5, 112, 182 MIT (Massachusetts Institute of Technol-
97, 137-40; endorsements, 138-40, program released on, 46-47; Berners-
ogy), Laboratory for Computer Science
148; IBM e-business mark, 137-38; Lee and, 15-16, 17, 23, 25-26; commer- Jobs, Steve, 23
Johnson, Tony, 64 (LCS), 54, 55, 60-61, 63, 72, 76, 127,
' standards for, 137 cial editors, 26, 44; Engelbart and, 6;
Electronic Data Interchange (EDI), 188 JPEG (Joint Photographic Experts 149; Berners-Lee at, 87-89; Consor-
free speech and, 139-41; global, 20, 22;
Electronic mail (e-mail): before WWW, Group), 165, 166 tium formed at, 81-89, 91-102;
Nelson and, 5; '91 conference, 50-51;
Junk on the Web, 134. See also Filtering web.mit.edu, 61; web server for W3C
18; confidentiality and authenticity philosophical aspects, 27-28; virtual
software (w3.org), 95
issues, 152 museum, 59; WWW and, 33-34, 183
Engelbart, Doug, 6, 50 Moffat, John, 206-7
Hypertext (1991 conference), 50-51. See
Enquire Within upon Everything, 1, 2 Kahle, Brewster, 40 Montulli, Lou, 68, 77
also European Conference on Hyper-
Enterprise Integration Technology (EIT), Kahn, Bob, 6 Mosaic, 67-71, 76-77, 82. See also
text Technology; European Conference
82 Kahn, Gilles, 102 Netscape
on Hypermedia Technology
European Conference on Hypermedia Keio University, Japan, 116 Mouse, 6
Hypertext Markup Language (HTML), 2,
Technology (ECHT 1994), 95-96 Kleinberg, Jon, 204 MUDDs, 172
29, 36, 41, 43-44; development of, 86;
European Conference on Hypertext Tech- fragmentation concerns, 98, 160-62; Klensin, John, 62
Knuth, Donald, 182 National Center for Supercomputing
nology (ECHT 1990), 26 metadata in, 181-82; reason for,
Kotok, Alan, 78, 109 Applications (NCSA), University of Illi-
European Union, 88, 95; Webcore, 79 40-41; rivals of, 41-42, 96; standard-
Krol, Ed, 76 nois, 68, 69, 70-71, 75, 76, 80
ization of, 54; tags, 41-42
Kunz, Paul, 45, 63 Navisoft Inc., 81, 105; Navipress, 81
Fermilab, 69 Hypertext Transfer Protocol (HTTP), 2,
Nelson, Ted, 5, 6, 64-66
Filtering software, 125, 134-35, 136 36, 38-39, 97; code written for, 28-29;
Legal Information Institute, Cornell, 69 Net Perceptions, 144
Fink, Steve, 78 format negotiation, 40; HTTP 1.1, 97;
Fluckiger, François, 88 Legal liability and the Web, 133-34 Netcheck Commerce Bureau, 137
https:, 150; standardization of, 54
243
 i n d e x
i n d e x

Netscape, 82-83, 92, 93; free release of, regulations, 146; hypertext protocols Separation of form from content, VAX/VMS, 15, 19, 50
83; IPO and stock, 106-7; Microsoft and, 141; P3P (Platform for Privacy 130-32, 168 Vezza, Albert, 82, 87-88, 91, 93-94, 95,
and, 91, 103; Mozilla (Navigator 1.0), Preferences Project), 147; P K C (public Shamir, Adi, 149 109
96, 99, 103; Navigator 2.0, 112; open key cryptography), 126, 149, 150-53; SMIL (Synchronized Multimedia Integra- Viola, 56-57, 63, 64, 68, 83
source policy, 118; security issues and software for, 147; Web site privacy pol- tion Language), 166-67 Virtual Library, 55
Secure Sockets Layer (SSL), 97, 150 icy, 146 Social machines, 172-75 V R M L (Virtual Reality Markup Lan-
Network Solutions, 128 Prodigy, 92, 105 Sociotechnical issues, 109-14, 124-41. guage), 86, 166
Newcastle University, 71-72 Prospero, 40 See also Censorship; Pornography; Pri-
Newman, Clifford, 40 Protocols, 18, 19, 36, 123-24; Document vacy; Protocols Web browsers: Amaya, 102, 119,
NeXT Inc., 22-23; computer, World- Object Model (DOM), 168; elements for Sollins, Karen, 54-55 170-71; Arena, 67-68, 98, 102; devel-
WideWeb code written on, 22-23, 28, the Web, list, 36; global systems, 35; Somm, Felix, 133-34 opment of first, line-mode, 30, 51; as
31, 45, 47, 4 8 4 9 , 55; failure of despite graphics formats, 165-67, 168; inde- Stallman, Richard, 73
r
editor, 32, 45, 57, 70-71, 81, 102, 112,
product, 28 pendence, 160, 168; Internet, 38, 40; Standard generalized markup language 169-72; Erwise, 56, 58, 59, 67; gopher,
Nielsen, Henrik Frystyk, 58, 97, 171 standardization of W W W s , 53-54, (SGML), 41-42, 43-44, 96, 119 " 40, 67, 72, 84; licensing fee contro-
NLS (oN Line System), 6 71-72, 98; universality needed, Stanford Linear Accelerator (SLAC), versy, 72-73, 75; Lynx (screen-mode),
NNTP, 38 163-65. See also Hypertext Transfer 45-46, 63, 64; first web server outside 68; Midas, 64, 67; Samba, 58, 67;
Protocol (HTTP); U R L (URI) of C E R N , 46 search for, 26-27, 32-33; ViolaWWW,
Online Privacy Alliance, 148 P3P (Platform for Privacy Preferences Stock market and WWW, 106-7, 124, 56, 57, 63-64, 67; Windows, Cello for,
Online research, 178-79 Project), 147 205 69-70, 77. See also Microsoft; Netscape
Online voting, 172 Putz, Steve, 59-60, 77 Sun Microsystems, 104 Web of Trust, 153-54, 193
Open source software, 171; community,
Web pages or sites: first, 29; privacy poli-
151 Quint, Vincent, 102 Tangle (program), 12-13 cies, 146-48
Operating systems, 19; communication TCP/IP, 18, 19, 20, 54 Web servers (HTTP): Apache, 120, 195;
between different, 17, 19 Raggett, Dave, 67-68, 86, 98, 116 Telnet server, 47-48 early classics, 59-60; first one,
O'Reilly Associates, 64, 76, 80, 96 Reed, Brian, 78 T X , 182
e
info.cern.ch, 29, 32, 44, 45, 49, 55, 66,
Owl Ltd., 26, 96 R E X X (programming language), 32 Thinking Machines, 40 75, 95; first outside of C E R N , 46; Jig-
Reynolds, Joyce, 54, 61 Thompson, Dave, 68 saw, 120-21, 170-71, 195; 1992,
Patents, 196-98 Rimm, Marty, 112-13 Totic, Alex, 77 spread of, 55, 59, 66; 1993, continued
Patrick, John, 112 Rimmer, Peggie, 14 Trust. See Web of Trust development of and Mosaic, 67-71, 79;
PCs (IBM and clones), 44, 77 Ritchie, Ian, 26, 96 Trust engine, 192-93 1994, Netscape started, 82-83; Telnet,
Pellow, Nicola, 29, 30, 42-43, 48, 58 Rivest, Ron, 149 47-48; virtual museum, 59
Pesce, Mark, 86 Rogers, Kevin, 8, 9, 11 Unisys, 165-66 Webcore, 79
PGP (Pretty Good Privacy), 152-53 RPC (remote procedure call) project at University of California, Berkeley, 56 Websoft, 83
PICS |Platform for Internet Content C E R N , 14, 17, 19; addressing scheme, University of Kansas, 68 Wei, Pei, 56-57, 63-64, 68, 77, 196
Selection), 113, 114, 125, 135, 136, 138 20 University of Minnesota, 40; gopher, 40, What Will Be (Dertouzos), 102
PRC (public key cryptography), 126, 149, 68, 84; licensing fee controversy, 72-73 Whole Earth Internet Catalog (Kroll), 76
150-53; government fear of loss of con- Saito, Nobuo, 116 University of Texas, San Antonio, Hyper- Wide Area Information Servers (WAIS),
trol and, 150-51; RSA, 149 Scheifler, Bob, 78, 82 text conference, demo of WWW, 1991, 40, 50
PKI (Public Key Infrastructure), 152 Screen scraping, 178 50-51 Williams, David, 22, 60
Plessey Telecommunications, 4 Search engines, 133, 177-78, 180 Unix, 17, 19, 44, 55, 76, 99 Wilson, Chris, 77
PNG (Portable Network Graphics), 166 Secret, Arthur, 55 U R L (URI), 2, 33, 36, 37, 39-40; code Windows: Cello for, 69-70; Mosaic for,
Pollerman, Bernd, 32 Secure MIME, 152 written for, 29; as fundamental innova- 76-77. See also Microsoft; Netscape
Poole, John, 11-12 Security issues: digital signature, 150; tion for the Web, 39; naming of, 62, Wollongong University, Australia, 19
Pordes, Ruth, 70 Secure Sockets Layer (SSL), 97, 150. See 127-29; prefixes, 40; standardization World Wide Web (WWW): body (consor-
Pornography on the Web, 112-14, 136; also Electronic commerce of, 54, 61-63 tium) to oversee formed, 75-76, 78,
liability, 133-34 Segal, Ben, 17, 23, 29, 48 " U.S. Defense Advanced Research Pro- 81-89, see also World Wide Web Con-
Portals, 124, 132 Semantic Web, 157-58, 177, 179, 181, jects Agency, 95 sortium; browser created, 1990, 30;
Postel, Jon, 127 182-83, 184-96, 205; inference lan- client, 28; code written for, 28-29; as
Privacy issues, 125-26, 143-55; click guages, 185, 188; knowledge represen- Van Dam, Andy, 27 collaborative medium, 57, 123-24,
stream information, 144; confidential- tation, 186-87 Varney, Christine, 148 157-58, 161-62, 169-72, 200, 206;
ity, 150; cookies, 145-46; European Sendall, Mike, 17, 22-23, 26, 46, 88 Vatton, Irene, 102 commercialism and, 107; companies,
244
245
i n d e x


releasing beta versions and giveaway browser, 119-20, 170-71; Apache web
software, 100-101; control of, con- server, 120, 185; Asian host, 116; Docu-
cerns, 124; damping mechanism ment Object Model (DOM), 168; evolv-
needed, 205; early steps in project, ability, 190; formed, 75-76, 78, 81-89;
27-28; fundamental principle, 37; ' global agreement, 188; government
future of, 190, 195, see also Semantic support for, 95; intercreativity, 169-72,
Web; gateways to VAX/VMS and WAIS, 201-7; internationalization activity,
50; global brain analogy, 204-5; global 167-68; Jigsaw server, 120-21, 195;
• growth, 57, 75, 80, 108-9, 200; hyper- Live Early Adoption and Demonstra-
text and, 16, 23; incompatibility tion (LEAD), 170-71; management
between computers solved, 35, 36-37; positions, 109; membership, 93-94,
infrastructure, 129-30; international 115-16, 118; network security issue,
conference, first (1994), 79-81, 85-89; 97; PICS (Platform for Internet Content
international conference |WWW6 Selection), 113, 114, 125, 135, 136, 138;
1997), 119; junk on, 134, see also Filter- policy and procedures, 109-10, 121;
ing software; as killer application, 33; P3P (Platform for Privacy Preferences
legal liability and, 133-34; as manage- Project), 147; public domain software
ment tool, 35, 174; named, 2, 23; neu- development, 121; purpose, 94,
trality of, 130-31; newsgroup to share 118-19, 189-90; Resource Description
information about, 47; on-line mailing Framework (RDF), 181, 182-83, 184,
list added (www-talk@info.cern.ch), 189, 190-91; sociotechnical issues,
50, 53, 171; open source policy, 171; 109-14, 124-41; vendor neutral status,
principle of minimal constraint, 39, 94-95; website (www.w3.org), 121,
124; protocols for, 36; program released 171; X M L development, 119
on Internet, 46-47; public domain sta-
www-talk@info.cern.ch (mailing list), 50,
tus, 73-74; self-regulation, 113-14,
53, 171
148; societal impact, 199-209; soft-
WWW Wizards Workshop, 76-77, 96
ware/commercial services start-ups, 83;
spreading of, 1991, 47-48, 49, 51; tele-
Xanadu, hypertext project, 5, 64-66
phone system analogy, 99; theme: inter-
X Consortium, 78, 82
play between political decision and
Xerox PARC, 54, 55, 63, 77
technical, 41; universality of, 163-65;
X H T M L , 162
vision for, 1-2, 12, 27, 31, 33, 76,
XML (Extensible Markup Language),
83-84, 91, 123, 157-75, 199-209
World Wide Web Consortium (W3C), 92, 119, 160-62, 166, 181, 184, 188
93; Accessibility Initiative, 167; Advi- XSL, 168 N
sory Committee, 97-99, 109; agree-
X Window system, 55-56, 67, 78, 99
ments about privacy, 147; Amaya
Yahoo!, 124




246
 •»




THE GENIUS WHO BROUGHT US WWW AND HTTP REVEALS
HOW HE INVENTED THE WEB, REFLECTS ON ITS IMPACT,
AND PREDICTS WHERE IT'S HEADED
Named one of the 100 greatest minds of the 20th century by Time magazine, Tim
Berners-Lee is responsible for one o f that century's most important advancements:
the world wide web. Now, this low-profile geniusâ€"who never personally profited
from his inventionâ€"offers a compelling portrait of his invention. He reveals the
Web's origins and the creation of the now ubiquitous http and www acronyms and
shares his views on such critical issues as censorship, privacy, the increasing power of
software companies, and the need to find the ideal balance between commercial and
social forces. He offers insights into the true nature of the Web, showing readers how
to use it to its fullest advantage. And he presents his own plan for the Web's future,
calling for the active support and participation of programmers, computer manufac-
turers, and social organizations to manage and maintain this valuable resource so that
it can remain a powerful force for social change and an outlet for individual creativity.


T I M BERNERS-LEE is currently director of the World Wide Web
Consortium, the coordinating body for Web development, and
he occupies the 3Com Founders chair at the M I T Laboratory
for Computer Science. The recipient o f numerous awards,
including a MacArthur Fellowship, he is based in Cambridge,
Massachusetts.



tm HarperBusiness ISBN 0-06-251587-X
An Imprint ofHarperCollinsPublishers
www.harpercollins.com
Cover design © 2000 by Mark Cohen
Author photograph © 2000 by Louis Fabian Bachrach


U S A $15.00 / C A N A D A $22.95