MIT/Brown Vannevar Bush Symposium - Fifty Years After 'As We May Think' (Part 2/5)

Search transcript...


MODERATOR: Well, it's a great pleasure and privilege to introduce another genuine pioneer, another person who's really helped shape our field in a multiplicity of ways. Bob Kahn was primarily responsible for the design of the ARPANET, which was the first practical realization of packet switching, which was rather a controversial thing at the time people first started advocating it. He was also responsible for initiating the internet project at APRA, which was the follow-on to ARPANET. And that was the first open architecture network, another very important idea. He's a co-inventor of the TCP/IP protocol, which provides the framework for the internet.

Now, that's one role in which he functioned as key designer/scientist/engineer. Another was that he was a enlightented funder of the ARPA academic community. And he was responsible for initiating and funding many star projects and stars. And we're all in his debt.

He's been a tireless champion of computers and communications and that interplay that is so important to us today. He left APRA for a very specific purpose, to found-- and he is now president of-- the corporation for National Research Initiatives, which he founded in '86 to provide leadership for the research and development that he felt would absolutely be vital for the national information infrastructure, the NII, or as it became later known, the information superhighway, or the infobahn, or whatever your favorite buzzword is, that that would take in order to be more than just hype to be reality and to be useful to all of us. And he is still engaged in working those research issues.

Now, bringing it back to our theme today, Bush had very little to say about how people work together, the idea of sharing, or as Ted now says, transclusion, was certainly in there, but it is almost hidden away. And you really had the sense that he was working on personal information management devices. But sharing information is obviously the name of the game today. And spools of annotated microfilm in the mail is not how we fortunately have to do things today. He simply did not anticipate computers and communications as it has grown up. And Bob's role here today will be to talk about ways in which to extend Bush's vision. Bob.

KAHN: Let's see. Would you have any serious objections if I took your laptop off of the podium here? The urge to press on these keys is too--

PRESENTER: Just can't resist a keyboard.

KAHN: Thank you. OK. Well, first of all, I am delighted to be here at this session honoring the 50 years of Bush's paper. And particularly happy to be on the program after both Ted Nelson and Doug Engelbart.

I have to tell you, though, that I think it would have been much nicer if you put them on in the reverse order because every time I listen to Doug, I feel nice, and smooth, and relaxed. And when I hear Ted, I feel like I need to lie down.


Well, just-- Doug reminds me of a cyber psychologist. I always feel like probing. And every time I listen to Ted, it's sort of a hybrid. And I keep going back and forth between Back to the Future and Saturday Night Live.

When I was asked to give this talk, Andy said, look, anything you'd like to talk about would be fine as long as it's in the context of Bush's paper because that's what this symposium is about. And I got to thinking about the real guidance I got because in listening to both Doug's talk and Ted's talk, I got the impression that the guidance from Andy to Doug may have been to only talk about ideas that had been well thought about for the last 40 years, in which, of course, Doug has been doing over and over again. And in listening to Ted, I had the impression that he said, don't talk about any idea that you've had in the last 35 years, so focus only on stuff that hasn't evolved since then.

To me, when I first suggested, well, what about the internet, and I can talk a little bit about packet switching, he says, no, no, that's retrograde. So talk about the future, look into what the future is likely to produce. And I said, well, then, how about something about the framework for digital object services, which is where things are currently at in my way of thinking. He says, no, that sounds too much like a computer science seminar. I really want you to speculate about the future, which I translated to mean, don't talk about any idea that you ever thought about before you show up on this podium.


And that's been a little bit hard to do. So I probably won't keep to the letter or that particular guidance. But I'll do my best.

Let's see. When you look at the paper that Bush wrote, and I've read it now many, many times, and on the one hand, it really is very prescient about some of the needs, and requirements, and desires, and ideals that we might have in the future. I'm particularly impressed about his notion of what reality is.

I mean, his reality is totally embedded in the inventions of the day or the near term. He talks a lot about photo cells and advanced photography. He talks about thermionic tubes. He talks about typewriters.

He's locked into the stuff of the day, as it were. He certainly doesn't talk about digital technology. And I want to get into that in the future.

He has a very strong focus in his particular writing on scientists and what's left for the scientific community to focus on. He certainly has pointed to the information overload problem, which certainly must be greater now than it was then. But it's nice to know that it's not a new problem. At least it gives us some hope if we haven't solved it in the last 50 years. We have some work to do in the next 50.

He talked a lot about the linking of ideas. But when you looked at the machinery that he was talking about dealing with it, it tended to be more grounded in logic than it was in higher level notions, more semantic contexts. And perhaps he really believed that somehow higher level ideas could be put into the context of logic as many people have proposed on certain axiomatic bases. And he, I think, emphasized very strongly the human dimension as he got into the notion of mnemonics and where it could lead through some of his ideas like trails and so forth.

But as I look at what he wrote about in the paper, I am also struck by what it was lacking as well as what it was prognosticating upon. It was really very strongly lacking in digital technology. And we heard even from some of the members of the audience that perhaps he was not all that comfortable with digital technology because it really arose from around him but not from within him personally.

It's interesting also that the paper was dated in the year in which-- or certainly in the decade in which computing first really emerged as a workable invention. Not to say that people had thought about it before and hadn't poked around with it, but I think the '40s was really the year of the early workable machines, in the '50s, the first really practical ones. And then after that, it really got commercialized seriously.

He was also therefore embedded in a world of analog devices. He was focused on where the science community went. And, of course, he was also well grounded in the physics community of the day, which, at the end of World War II, saw its mission coming to an end potentially.

I think we have some of the same similarities today, but a very different possible rollout. Just as the end of World War II offered serious questions and challenges for the scientific community, so to today does the end of the Cold War offer some serious challenges for the entire scientific research community in this country because we have come to the end of what's been a very long 50-year compact between the federal government and the research community in the sense that a lot of the research that was supported over that period was predicated on the need to do it because of the Cold War.

And while that's by no means a suggestion that the federal government won't continue a compact with the research community, it's pretty clear to me that it's being reformulated as we speak. And nobody yet knows what the shape and form of it is. I'm particularly hopeful that it will come out with as strong support as we've had in the past. But it's by no means guaranteed. So, there are some parallels between what we've seen in the past and what we're seeing now.

The fortunate thing in Vannevar Bush's time was the fact that he had this future that was unveiling itself as he was writing his work. Today you could say that some of the same is happening. I could argue that we have not yet really seen the innovation that's going to really change the world of the future. Maybe we have it in our heads, and it hasn't come out. Or maybe it's happening around us, and some of us don't know about it.

The absence of digital technology in his vision it seems to me led to a very limited view of what was doable and what was not doable. In 1972, when I played at the very first ICCC conference a tape about the ARPANET and the origins of packet switching, a lot of the people who were very active and have stayed active in the field were on that tape. And had we done a retrospective, I would have played it for you here today in part.

If you listen to that tape, what you will hear is that the communication folks speak in very timeless phrases. They talk about the need for connectivity, high bandwidth, highly reliable systems. And you could play it today and say it's today's challenge just as well.

When you listen to the folks who are more grounded on the computer side of the world, and I think the distinction was larger then than it is today, almost everyone was talking about the technology of the day. And so almost nothing that they said really translates easily into today's context because it's dated. So you would see people talking about the need to connect a high power graphics display to a powerful remote time sharing system as their aspiration for the future.

Today, nobody would argue that. Nobody talked about much of the technology that showed up because it just didn't exist. They talked about better use of the technology that they then had. But they were very clear about the role of software in this whole thing.

Well, where are we today? I say still in very embryonic state. We are rapidly converting to digital. But lest you think that we've made a lot of progress to date, we are barely there.

Today, if you can imagine a not too distant future when we might have, let's say, upwards of 60 million people on the internet, keep in mind that that's likely to be less than 1/10 of 1% of the world's population. So that the actual size of the population that we're-- or maybe it's 1%, but the size of the population that we have to deal with is much, much larger than people who are now experiencing this particular technology.

And as the technology gets out, the biggest problem I think we're going to have to deal with is the cultural distinctions between the network itself, and what it implies, and the physical reality that we all deal with on a daily basis. So let me separate these two things and talk about the physical reality and the network and give you some notion of the kind of cultural disparity that exists between these two worlds. I mean, I've experienced myself in a number of different contexts. It's come up in some of the dealings with the financial community when they have largely in the past-- and most people think about using networks as a vehicle for moving information about money.

What they don't realize is that you can also have a notion of money that exists wholly within the network itself, has no necessary physical counterpart in the outside. So, you open up your wallet, you take out a string of bits, you hand it to somebody else, and you've done the same thing as take out a dollar bill and hand it to them. That is, there need be no physical account anywhere else in the world that knows about that dollar bill or knows about these bits. And yet, you can pass them around, maybe even make change for them.

Well, when you describe that to people who are somewhat sophisticated about digital technology, very first thing they think about is you must mean passing information about it no matter what words you use. It's very hard to think about anything else.

Second example I would give you is in the working that we've done with the library community, where the library community traditionally has been very paper-based. They deal with books. They deal with stacks of material. And their notion of the network is it's a way of taking that paper world and dealing with that information in a more flexible-- flexible way. And so what they tend to do is take their thoughts and their approaches that were well developed in the paper world and try and bring them into the world of the network, oftentimes at the loss of all the new flexibility that you could get if you rethought it completely.

Anyone who has ever tried to take a book and put a book on the network in anything more than just showing the pages one after another or letting you move through them more flexibly realizes that the design of something for the network world can be a very different challenge than a design of something for the paper world. And I'm not talking about just the ability to do hypertext linking and things like that. The very nature of that experience can be very different. So the heart of the problem as I see it in this particular cultural context is getting people to understand that when you're in the network, you may be having a different experience than when you're outside.

And so there's the whole idea of where you're grounded. And when you're grounded in reality, the network is a tool. When you're grounded in the network, then reality becomes your tool, as it were. You have things flip around completely. So this shows up in a number of different domains. But let me just start out by talking about creativity in the digital world for starters.

People that we deal with in the various industries that are content-oriented today know that when they produce a work in paper form or if they record something, it has a reality in the physical world. And when it gets translated into the digital world, they want to think of it as being the same thing. A lot of laws that apply in the physical world that they'd like to apply in the electronic world, and maybe that's the right way for it to come out politically. It doesn't have to come out that way technologically.

There are people who can create things in the digital world. That is to say if you are a creator of a musical work, there's no need necessarily for that work to first show up in audio recorded form before it ever gets transduced into any way that's captured and stored in memory. That is, you can create a totally in digital form without anybody ever having heard it before. And in that sense, it's got different properties than it might if it were in the normal analog physical world.

The act of converting something from the physical world to the digital world is to some a trivial, mechanical contrivance. It's just a matter of scanning some bits, digitizing some bits. But what you end up with is something that is uniquely different in many ways.

It's manipulable in different ways. It's processable. It's transportable. There are things that you can potentially do with it. And in fact, it may be a very new contrivance all on its own.

I'm reminded of the discussions that I've had with my wife, who happens to be a copyright lawyer, about the history of some of the key developments in this field. And I was particularly struck with how difficult it was to get people to realize that a motion picture was really a different kind of thing than just a series of images.

And yet while in this country we recognize the existence of the motion pictures, audio visual work of a different kind than just a collection of images, because it does have a storyline, and it does convey a different experience, there are countries around the world for which motion pictures are protected as just a very long series of images. And there is no other attribution given to them beyond that.

So, one of the hypotheses I want to put out is that, in general-- and this may not be true in every single case, and this is certainly not a statement about how laws should be written-- but the fact of the matter is that digital representations of things are in fact literary works. They are putting together things in terms of sequences of symbols, which is what a literary work is. We tend to put them together as sequences of the alphabet, which are non-processable all except by a human brain and system looking at it.

Potentially English could be processable by a machine in the future according to some rules. But when you write computer programs, and you're executing on digital representations, you in fact are dealing with literary works in a program field according to my reading of the law today. Whether that is something that would be universally accepted or not is a different issue. But it seems to be the case.

Now, when you talk about a digital representation, the interesting word there is not so much digital but representation because it implies that what you have inside the system is a representation of what was outside the system. And what do you do when you start out with something that's inside the system to begin with?

Well, it's not clear it's a digital representation of anything else. It may be the first and only fixation. It may have an analog representation of the digital original, if you will, but it may not be a digital representation of something that was outside like a digital representation of an image, or a movie, or whatever.

Furthermore, as we move to an all-digital world, my assertion is that our grounding in reality is going to be lost in a very fundamental way in the sense that as we get into more and more activities that involve multiple participants enabled through this digital medium, everyone will be able to be involved and participate, witness, view, experience a given event or set of events in their own unique way.

And you might say we do that today all by ourselves. And an arbitrary camera, though, could record the facts as they were. And so the facts of the particular setting could be recorded on film. And we would say that was the reality.

But in a digital space, maybe you can't get to a single place that was the thing that produced the experience that everybody had. And since they all saw it differently, the Rashomon phenomena are likely to abound in multiple ways. So there may in fact be no real reality anymore, just multiple perceptions of the same thing coming together.

Well, when you look at Bush's notion, and you realize that digital technology didn't fit in his world, you realize that he was pushing pretty hard, and in a rather remarkable way, to figure out how all these analog pieces of technology could play into this future. But when we look at it today, that range of technology almost is not on the visibility scene. We have semiconductors and electronics more generally.

My guess is that in time even that technology may be superseded by something else. I mean, we're all looking at optics and photonics. But that may not be the last word in terms of materials and material phenomena. In fact, as we move to faster and faster and smaller and smaller, we may get from our current views where we're talking about moving things to actually manipulating things in smaller and smaller worlds as we get into the ability to manipulate the nanoscopic and smaller.

On the networking front, the issue is somewhat similar to computing in the sense that I think computing as we know it as a machine that's able to manipulate symbols on our behalf will be around probably forever. Networking in terms of its ability to provide reliable connectivity between different parties will probably be around with us forever. I don't think those are likely to be taken for granted anytime soon. Maybe in the future it will.

And so when you look at what is possible within this framework, begin to realize that while many people are pushing the technologies, and we get amazing developments almost every year, in fact, the real action is likely to be in the systems and the application side of the world where we will try and put these things together in interesting ways that take real advantage of the power of the networking and the power of the computing capabilities, both the equipment themselves and all of the peripheral devices that go along with it.

Everybody here probably would view progress in networking in terms of a few simple measures like bandwidth, moving from 56 kilobits or 300 bod for those old timers to one and a 1/2 megabits, or a gigabit, or a terabit or a petabit. And I run out of Greek prefixes at some point here. Somebody could probably help.

On the computing side, software is still too hard. But the real challenge is putting it together. It's the integration function.

And as you think about the integration function, you realize that the idea of integration ultimately has going to be more than just plug and play. That is, plug and play presumes that somebody has figured out what the architecture is for coherence beforehand and put out this technology in a form that you can just plug it and things will work.

But this is unlikely to be the way things actually work. It's unlikely to be the way things work at the technology level. It's unlikely to be the way things work at certain dimensions of the human interaction level. And it's unlikely to work that way at the content level.

Now, if I could have the first of the slides. I just want to give you the NII example of how this integration can and cannot play out. There are a lot of people who think that the NII, so-called National Information Infrastructure is a kind of thing that we need. And we might consider putting out an RFP to somebody to go build it, like a building. So it's got an architecture that's below it. But in fact, it's probably more organic than that.

To really be effective, it needs to have the kind of dynamism that the economy might have where you can capitalize on ideas from different sources, put them together so that the whole is greater than just the sum of the parts. But this slide illustrates one model of an NII that might come out of our combined activities. But what is this? It's a very uninteresting notion of an NII. It is more like what we would define the economy to be, namely the set of all activities that are in [INAUDIBLE] production and consumption, social and otherwise, in our system.

Here's the same thing for the NII. You simply take the collection of technologies, and applications, and all the things that people produce, and all the things that people consume, put a circle around that, and call that the NII. If that's what we end up with, we will have failed to leverage the real possibilities of this revolution in computers and communications that we're experiencing.

The infrastructure is absolutely crucial. And the reason it's crucial is that if it's done right, it can lower the barriers to productivity in lots of different ways. You've heard some examples already from some of our speakers.

But we haven't really had a national-level infrastructure despite the fact that we've had a telephone system in this country for a long time. I heard one of my colleagues say this was a hundred-year mistake in going from all digital technology in the telegraph to all analog technology and now back again. But that's another matter.

We may, in fact, go through this cycle again and again as we consider whether to move to an all-optical plan, which is not yet clearly digitized. So we may move back to an all optical one at some point in the future only to realize the power of digital technology in the network, and then figure out how to gain the power of that through some physical invention of which I can barely begin to comprehend at this point. So, you have this conundrum.

If I could have the next slide. This is the dream that many people have, that you take all of these different capabilities that are going to be part of the NII, and they go through a magic convergence process. Don't ask me what it is. I saw a wonderful film on the notion, which made fun at this whole notion as to whether it's possible.

But I would describe the first slide as being akin to putting out an RFP for a symphony asking for violin, oboe, first cello pieces and saying that there can be no more than 30 minutes in length and have a dynamic range of 60 dB. And then you take the-- you take the ones that look nicest on the screen, you add them together, and you call that your symphony.

The problem is it's missing something. And I know what to call it exactly. It's not the architecture of the symphony.

But it is the musical work that was presumably in the symphony in the first place. So you're not going to get that musical work by linear addition. It isn't going to happen, except that unless the monkeys-- enough monkeys submit musical compositions.

So how is it going to happen? Well, one way it could happen is if magically Mozart could figure out what the NII symphonies are for each of these different areas and come out with the right prescriptions for what the oboes, and the flutes, and so on should be playing. But the NII is so big and broad that Mozart needs a pretty dynamic range to be able to cover all these bases. And I don't know anybody who's broad enough to cover absolutely every one of the bases. And maybe there are some who can cover some of them. But we don't even know what all the bases are.

So how are we going to get to this convergence? And more importantly, how are we going to get anybody to agree to this notion of convergence? I think it's going to be very hot.

I want to come back to this because the key to this working is open architecture in my view. And it seems to me that the notion of open architecture is not necessarily something that is going to be unanimously agreed to a priori. We may agree to an open architecture a posteriori once all the standards, and interfaces, and the like are in the marketplace.

And if I can have the next slide. The more likely scenario is that we will have some degree of convergence in some limited domains initially. And that we'll have a lot of separate efforts by different groups, some with vertical integration and some not. And sum of those two things, in fact, will be what we decide to call the NII.

Now, as you think about the open architecture issue, let me try and describe this in internet terms. When I was first working on the internet architecture, I described it to a large number of people as an open architecture approach to networking. It basically says, anybody can bring their own notion of a network.

You can plug it together. And somehow, the whole thing will seamlessly work to do best efforts delivery of packets from source to destination. And it doesn't matter what the network technology is. Furthermore, you can have value-added service providers come and provide their own networks. In fact, the whole thing could be commercialized at some point and still allow for evolution of the technology.

Well, in the early '70s, we had one large communications company in this country. And the idea of them building an open architecture network for lots of other people to participate wasn't exactly in their game plan at the time. And in fact, I suspect broadly speaking, even for people who would have been open to that idea, it didn't mean anything to them. They didn't know how to relate to it.

I could see similar parallels to people trying to relate to hypertext when their only model of the world was a book. But when you're trying to relate to a network, you think about how you put it together, how's it going to work? Well, what's the issue here? Nobody's really in charge. The framework is responsible for keeping it working. There might be a process through which evolution can take place, but nobody really is running it.

It just didn't have the ring of truth. Nothing this important to our society should be run that way, except the economy maybe or a few other things like that. So it just didn't have the ring of truth back in those days to anybody who didn't understand it.

So where are we today? You take the area of electronic transactions, which libraries will fall into, and electronic commerce would fall into it, all of those things, and you say, what we really need is an open architecture for digital libraries, let's say. And everybody says, yeah, that's fine.

But then they glaze over because it doesn't mean anything to them. They don't know what does it mean? And I say, well, it's really simple. You bring your own library to the thing, and it seamlessly becomes integrated in the whole.

You say, yeah, great idea. And you go on to your next thing because it doesn't translate into a technological plan of action. And even if it did, it doesn't translate into a political, systematic, commercial plan of action either. At least that's the way most people would look at it.

So, you go to someone who is in the database world and say, this is what we want to say, and we have the answer. It's called distributed database technology. That's what it does. Except it's my distributed database technology.

And you say, well, I want to take your distributed data based technology and make that an open architecture. And you say, wait a minute, that's where our value-added stuff comes in. And we can't possibly do that and make it available to anybody else. And I'm going to come back to this in the context of content in the middle because there is an analogous notion of an open architecture for dealing with content. But let me do that in due course.

So, here you have this issue of open architecture, which I think is important. But you have to realize that there are many people in industry who would look at open architecture and say, that's what we get after the whole process of sorting out the marketplace is over. And if you try and do it upfront, it precludes any single industrial giant who thinks they can do the whole vertical slice from weighing in and taking over the market.

So some people think it's a good idea. And some people think it's a bad idea. But I think it's the only way you can get this organic growth into a very large system where everybody has a chance to participate.

Well, then you think about organic growth, and that leads you immediately into issues of adaptability. If I could have the next slide. I've got some of these points I'm going to touch on outlined here.

So I've already talked about digital technology. I'm going to come back to a couple of the points there. But I'm jumping down right now into adaptive behavior.

The most important thing you want in any system-- and again, I'm thinking of network in the broad sense of including not only the communications and all the different pathways, but also the machines that may be connected to them. And when I talk about communications, I don't only mean terrestrial lines. I do mean wireless, over the air. I do mean all of the different media that one might use in order to get information from one place to another, electronic, optic, or whatever.

And I would even include in this open architecture the possibility of non-permanent connections so that people can in fact exit from the system and plug back in and still keep things going. So you'd like this kind of system to be able to adapt itself. And ideally, you'd like it to be able to adapt itself in some intelligent ways.

How many of you know of any systems that really have a detailed awareness of what it is that they're doing? And the answer is almost none. I mean, they might know, for example, in the phone company how much loading is on their lines, or they may keep some traffic statistics. But in terms of what's really going on, they really don't have a notion nor any idea of how they can then configured the net to do a better job of dealing with this.

To give you an example of how you could get a better idea of awareness or even learning, let me hypothesize that in the world of the future, which might lead to very, very complex functionality or might not, that individual components in the net may have the ability to know what it is that's going on. A given wafer, for example, might no longer be just a set of electronics carrying out a function, but could, in fact, be multiple levels where level one, in one example, might be carrying out the function, and level two might be overseeing the function carried out by the other level. Now, if something goes wrong at one level, the other one might know about it, may be able to fix it, and so forth.

Likewise, you can have synergistic implementations where both levels are providing functionality and also at the same time keeping track of the others. And that's only a two-level model. You can have trivial models that have more than two levels dealing with just replication and just checking to make sure how many are doing the same thing. But that's not really very smart. I think we can do better than that. So I think the idea of incorporating adaptive behavior in the net is an important one.

Another thing that becomes important-- I'm going to come back to this later-- is the possibility of injecting intelligence into the net. That is to say, today we're living in a world where everything that happens inside the system is principally controlled by the person who built it, brought it to you, runs it, operates it. There are very few cases where the intrinsic operation of that system is allowed to be controlled by arbitrary people from outside sending in the equivalent of, as it were, intelligent agents in the system that could effect its performance.

This smacks to many people of allowing viruses to come in. And it raises all the issues of trust and control. It raises all the issues of scoping and so forth. But it seems to me it's an our-- it's in our future.

Before I get off of this slide, let me jump up a couple of levels and just touch on a few other points I didn't make on this slide. One is that Bush in his article talked a lot about capturing information from everywhere. I mean, he talked about, I don't know, walnuts on your forehead-- I forget exactly how he put it-- microfilm just snapping pictures everywhere, hidden shutters in your pocket. The fact of the matter is we are rapidly getting to the point where every one of us can make digital fixations of everything that happens in our daily lives cost effectively in the predictable future.

And the real question is, what are we going to do with all of this information? We talk about information overload, this also becomes our intellectual property, a very fundamental nature, and raises some very profound social and cultural problems for the country. Every one of us has the potential for other people wanting to see as much of this information as we're willing to make available.

At some time in the not too distant future, there will be the first woman president of the United States. And that person may or may not be a little girl right now. And everything that goes on in that person's life may be of interest to the country as a whole. But you won't know about it until that event happens.

And so seems to me that if we were in that state, we ought to be making it possible to retain as much of this information as individuals are willing to allow to be retained, subject to their administrative controls, so that we do not lose our culture in the future. That's what I meant by digital fixations. Also, the ability to use this technology to allow collaboration and teamwork.

I'm going to come back to this in a few minutes. But there is a very distinct difference in view of what networking can do depending on what business you're in. Here's one view of the network as the medium for production and consumption of, let me say, information, maybe it's content. That is the idea if somebody produces something of value, the network is the delivery vehicle. Somebody buys it, you pay for it, transaction is done. This is trying to simulate the buying and selling of the economy in the network environment.

There's another model of the network where it's a medium for collaboration. It enables things to happen that couldn't otherwise happen for which there is no equivalent counterpart in speed and power in the physical world. And it seems to me that as we develop these ideas more fully, we're going to see these basic assumptions about what the network is having to come to the surface so that we can deal with them, both in law and in terms of the whole evolution of these systems.

In the notion of in cyberspace that I mentioned earlier, I think it's going to be possible for all of us to actually have a digital presence in cyberspace. Today almost all of us relate to it in one way or another. And that relationship is varying at different rates for many of us.

When I first started using the network in a serious way as a tool, I found myself initially going back to pencil and paper because I was comfortable in that mode. Because using the machine, I had to remember too many things. Too many things could go wrong. I didn't either quite trust it or understand it.

Well, I didn't feel relaxed enough to just relate to it in that mode. And then somewhere along the line, it just became the natural way to do things. And then the paper world became the arcane way. And for many people, I think they're going to discover a presence in cyberspace that's different from our current notion of just relating to it as a place where you go to pull things off a web or where you go to send a piece of email, but you actually have an existence there that's every bit as important in some ways as your existence outside of that space.

On the area of open architecture, again, the issue here is a very profound one for the scientific community. Because historically, the development of science has been predicated on the notion of individuals doing the work and therefore publishing their work in a form that others can review it. But when you're trying to actually create infrastructure, oftentimes the full resolution of a problem and all its details is an inhibitor to getting other people to adopt it.

That is, the very fact that you stipulate an approach when there might be 1,000 different approaches is an inhibitor rather than a solution for actually creating the infrastructure. It may be an interesting intellectual idea, it may be a contribution to knowledge, but it may be an inhibitor to the actual creation of infrastructure. So the hard part about infrastructure is excising all of the details for which there may be lots and lots of different and alternative approaches that are better than any single one that you might come up with, or where you don't even know that the one that you have maybe the best, allowing for people to participate in that process. If I can have the next slide.

I think that the real issue in network enhancement is, one, what is the network itself? And number two, what kind of functionality are we willing to allow others to stipulate for the network? In the case of almost every situation that I've dealt with, people have shied away from allowing other people to have a key part in describing functionality. They're willing to take inputs, but they're not willing to let them take actions that produce that functionality.

And so I think this is going to be particularly interesting as we get into issues of learning in the network. And by learning, I don't necessarily mean that to have a scholarly notion as one might normally attribute to that term, but rather ones that can incrementally update themselves, they can figure out what's going on, that can detect patterns of behavior, do things to improve things that can be codified in the form of computer programs that can cause specific actions to take place. That's what I mean by my that term.

Let me go into this issue of this ubiquitous content capture, which I mentioned briefly before. If we have all of this information that's around, and we have it all in digital form, then we all become intellectual property purveyors in the most profound way. I suspect that a few of us in this audience have become aware of the value of intellectual property either by virtue of their selling videotapes, or by virtue of turning things into books for sale, or what have you.

But when you actually have all of this information about yourself, you don't know how much of it is going to have value. Therefore, you probably want to protect as much of it as you can from misuse by others. And this gets into a topic that I want to spend most of the rest of the time on. And that is the issue of malleability.

Once you've got a system that's capturing all of this information, it's going to be very easy to keep track of a variety of things. People have noted that in the past.

Bush's paper was particularly cogent on this particular idea that you could settle things up, you could keep track of schedules, building links. And the digital technology makes that possible. This is one of the areas where I think the predictions are more likely to come true. And they can be in various forms. They can deal with money resolutions. They can deal with logistics, your schedules, what have you.

And furthermore, the technology is all going to probably become more integrated with us. I mean, I can see the day when the threads in my jacket will be mini antennas. And if you want to write on your display, you'll just make a little note in your inside of your jacket pocket. And I mean, those notions of the risks, whatever, probably will come true in some way, shape, or form because that's just progress in technology.

But if I can have the next slide, I want to talk about this issue of malleable content. This is actually a very fundamental kind of issue. For anybody who has looked at the way copyright is handled today, you realize there are essentially these two distinct notions of the network, the network as the production/consumption medium for content where the idea is that you can buy it, you can have it, you can use it for your own personal use, but you can't really change it.

Now, there are people who are willing to put material out for use on the net, and they're willing to allow people to change it. And this is not to say that much of today's technology can't be adapted to it, but the people for whom this is a business generally don't want that. And so this is what I would call hard copyright protection.

The extreme other side of the coin is what I call "no fixation interchanges." I mean, this is the dilemma of the telephone versus the telegraph in the early days when people thought that the telephone wouldn't fly because it didn't give you any hard copy. And therefore, it was viewed by some as not being that relevant, certainly for business transactions.

But the fact of the matter is that in between those two extremes where, in the case of a telephone call, there is no work, it's just an interaction between two parties that never gets fixed, and the other side where somebody actually creates something that they say you can have, pay me, don't change it, there is a whole world in between which I want to call the world of "malleable content." And I think this is where the real ultimate value of the network is going to be played out because it's a world that we haven't experienced before in any real sense.

Now, let me just say a little bit about what I mean by malleable content. What I don't mean is that you take somebody else's novel and completely change it, or somebody else's treatise on a subject and create the inverse. Although, it's possible to enable that to happen as well. What I really mean is where somebody can put out a piece of a work into the cyber world, and as a result of that, other people can then build on that work in predictable ways where people get credit for what they've done, and where the intent is to have other people to malleably manipulate that work.

Now, this can be done in a number of different ways. But I'm particularly impressed, for example, in the discussions about education where people think that the task of providing educational content is providing things on the screen that you can look at like papers from a book. Whereas in fact, for kids to really find that interesting, they're going to have to interact with that medium in a way that's exciting to them, just like as if it were a nonfictional video game of some sort. So it's the interaction, the ability to cause something to happen as a result of an action that you've taken or a work that you've created that you add onto the other one.

It's possible that as we develop knowledge structures for use in various areas as time goes on that people who develop new knowledge in those areas can add that knowledge to existing bodies of knowledge, therefore making the knowledge malleable with time as we learn more about it. The interesting thing about that is if somebody adds some piece of knowledge which makes another body of knowledge much more valuable, it may be that the real contribution has no meaning whatsoever without the previous body of knowledge that came before it. And so you can't really packaging them and sell them separately. They are intricately entwined.

This is not to say that we haven't had experience with this kind of thing. But generally, the experience that we have had is where people have put out works of software, literary works of software-- and I don't mean that necessarily as words, but I mean symbols in digital form-- where the intent is for other people to use them and build on top of them. They may have certain structures that don't let you get into the innards of them, like operating systems that you can't change but you can use, or other programs that have the same or similar form, but it seems to me this whole idea of creating malleable content out in the net is the thing that's going to make this whole field really take off because it empowers people to become part of the process and contribute.

To do that, I think we need to back away from some of our current economic notions of how to pay for this. Because if the real value of downstream contributions depends upon licensing at all stages of the game, it's probably going to die in legal contracting world and in cyberspace. The fact of the matter is that we ought to have a mechanism whereby the action of a 28-year-old in Singapore need not bring down an enterprise just because they've distributed something. In fact, if 40 million or 400 million people actually put their hands on this material, it ought to be that the party who generated it in the first place is reimbursed more significantly for having done that. So I'd like to see that happen.

In the context of the last two slides, I think the whole idea of putting intelligence in the net is going to depend upon getting standard interfaces into these nets. I foresee a day when in fact we literally will have a single means of interfacing, one standard interface, that lets all of the functionality of that interface be reflected in the object that passes through it so that the issues of plugging and the issues of whatever are really grounded more in the fundamentals of what you're trying to achieve at the systems level than they are at the physical hardware or subsystems level.

If we can do that, then I think it allows us to deal with entities that can move around fairly dynamically, can use different languages, can do translation because people can inject into the systems the interfaces that are needed to make this all work. Now, we may not call that intelligence in the sense of traditional definitions, but I think it makes these networks more malleable in their own right and more functional in their own right in ways that we can all benefit from.

Finally, let me just say a few words about notions of really complex systems, which is where we're moving toward. I think that it is going to be the case before too long, if it isn't already, where no single individual really has any notion of the size or scope of what's really out there, what's connected to what, how anything is working anymore, just notions of the base that you're building on and how to add new functionality into it.

And I think that as a result of that, we're going to have some serious concerns about security, we're going to have some serious concerns about performance and functionality, and, in fact, systems collapsing of their own weight without any notion of understanding where the weight distribution is in these complex systems.

So I think we're going to need as part of this matter of developing this intelligence in the net ways for these systems to describe what they're about so that people can ask them and those systems can then tell you what they're about, and how they're structured, where their strong points and weak points are, and so that we have a better idea from the systems themselves as to what's actually going on.

There was a fellow that I used to work for who I think was the last person who actually knew all of the detailed implementation in the Bell System. And when he retired, he said his fear was that the whole thing would possibly collapse if nobody else could really keep their-- keep the whole thing in their head. But, of course, that really didn't happen as a technology thing. It happened as a political thing.

And that's probably the more likely event here. That is, we'll get to the point where the systems in fact can explain all of this stuff, but it'll be so complicated for people to then understand it that they'll decide that we need to rebuild, remodularize, and somehow restructure just so that they can better understand it. So the ultimate challenge, I think, for us is to have people understand what's happening who are not technologists but who are either users of these systems or ones who depend upon it in some fundamental way.

I am sure that if Bush had lived in a different world and he had different technology, he would have envisioned a lot of these things. These are not unique ideas. These are not things that would have been terribly surprising to him. It's just that when you're a very practical person, as I suspect Vannevar Bush was, that he tended to rely on what was then and what he could see in the likely-- in the likely prognosis for the near term.

And in closing, I would just point out that it was very-- it was very interesting to me to observe how in the short term almost everyone I know is more optimistic about what's going to happen and really does. But in the long term, we're all much more conservative about what's going to happen than really does. And I think this is somehow related to that old adage that if you lose a little on each piece and sell it in quantity, you'll make it up in volume.

So just how that happens or why it happens, I do not know. But it is a very strong curiosity of our field that we are way over expecting in the short term and way under predictive in the long term. Thank you very much.


MODERATOR: Stay here. Stay here.

AUDIENCE: I wish I could afford to have that perspective. But my perspective has to factor in cost. And I think that's borne out by the fact that you said that only maybe 60 million people are on the internet worldwide. And that's 1% of the world population, whatever, whatever.

There are people like John Perry Barlow who want to basically climb into their computers and live in cyberspace. And as this world that we're building becomes more attractive to the people who are there, the tendency for those guys to do that probably will increase. And so what I'd like you to comment on is what's it going to take in terms of cost drivers to get the 60 million to turn into 600 million or a billion before we end up with the interesting societal pressures of the network haves and the network have nots?

KAHN: I mean, I certainly do not view this as the only metric of networking bandwidth or even connection bandwidth. I was giving that as an example how some people measure progress in networking. And clearly, their progress is a function of cost and their ability to afford that.

I think that we are going to have to move to a service-based notion of pricing before this will ever happen. Because as long as the vested interests are focused on existing revenue streams, you don't want to do anything to jeopardize that. I mean, the thing that has been so profound in my view about this technology development, and I guess you could say for others that have the similar characteristic, is that they're often brought to you by the antithesis of the folks you would think would bring it to you because it really is challenging the status quo in many ways.

John Mayo gave one of the keynote addresses at the National Academy meeting in Washington last year. And he said, it's a true revolution in the telecommunications industry. And it's only recently been realized.

And somehow this notion of the internet has grown up around all of these companies. They don't quite know how it got there or what happened. But it developed around them. And now they're trying to adapt to it. And my guess is that if they're at all smart, they will try and be the provider of the service for it. It wasn't something they had projected in the first place because no one of them would have probably taken the lead in building such an open architecture system.

If there are people who can in fact afford to provide connections at low enough price, then my guess is you will see them doing that only because it's their way of getting a leg up in a field that they otherwise couldn't compete for. And they will force other people to have to deal with them. And once you get the cost of this down to the commodities level, it becomes very hard for third-party folks to get in. But then you're dealing with base costs that are probably about as low as they're going to get.

My expectation is that what we're going to see is the tall pole in the tent grow taller. That is to say, we'll have more fabric on the tent. And you'll have a bigger base, and you'll have a higher tall pole, which means that the people who are working at the forefront will be further removed from the people who are at the base, and that you'll have lots more people at each of the levels, and that people will, in fact, migrate from capability to capability as they change jobs, as companies decide to make bigger investments, and that we will forever have a schism of capabilities as people will choose to buy more and more capability as they can-- as they can afford it.

But that the price for getting in at the low level will be reasonably affordable. This is a topic that we've dealt with on the President's Advisory Committee a lot, universal access, not necessarily universal service. And I know it's a topic that's been written about fairly extensively for those of you have looked at the black and white creatures and in snow crash would know.

MODERATOR: I was just told by Jim Rose that we have a very lively and pretty large group in the outside world that is following our proceedings via MBone and Paul Penfield's last-minute inspiration, which is his companion system for letting folks on the web page through the slides as each speaker speaks to them on the screen here. And we've decided on the spur of the moment that we'll allow some people to come in with questions for the panel using the whiteboard facilities of the MBone. So it's nice to know that we're not the only people taking advantage of all these great speakers.

Speaking about great speakers, I'm about to introduce Tim Berners-Lee. Outside of saying he's Mr. WWW, else need I say? Some think that the natural progression in augmentation tools, to use Doug's term, is the invention of writing, then the invention of the printing press, and now the web. I submit this may be a bit of an exaggeration, but it's certainly the case that the explosive growth of the web is truly an amazing techno-sociological phenomena. There are a few things that we can compare it to.

Tim has been and continues to be the prime mover behind this explosion. And I'm really happy that he's going to be able to talk to us today about his most recent thoughts about the net and where it's going and where the revolution leads next. Tim.

BERNERS-LEE: Thanks Andy. Hey, was that what you said? What a day. I mean, this is a-- Andy, this is great. I feel tremendously honored to be involved in it. I've got a hard job coming up here and talking for 3/4 of an hour following the gurus of hypertext. I feel a little bit like a designer of a steam engine following off with Boyle, and Charles, and Watts sitting in the audience.

I feel the same as Bob, actually. When I read over the Atlantic Monthly article again, you're so struck-- when you do that, you're struck with two things. You're struck by this incredible vision, the incredible long distance vision.

And you're struck by the fact that the guy was on the verge of the digital computer revolution. And he didn't have awareness of that. And he still produced something which is really very similar to the things we're trying to make now even though he didn't have the benefit of that awareness of the digital computer revolution.

It's interesting for me to look also at where he's coming from in the sense of the problem that he's trying to solve. That's interesting because when you look at the contributions that people have made to the field of hypertext and networking, that they've come often from different-- solving different problems. They've come from education.

They've come-- for example, Doug Engelbart with [INAUDIBLE] and, to that extent, I with the web, we're trying to solve the problem of people working in teams. And Ted coining the word hypertext very much from the literary standpoint, literary machines-- I've got my Literary Machines book here autographed, Ted Nelson, 92-8-24, which is, in fact, the first place I read a the "Mnemonics" article.

So, looking at the "Mnemonics" article, the problem that he identifies is that of an investigator of a scientific researcher. Can we have to first slide projected, video on? Great.

This investigator-- so the problem he sees is of a scientific researcher. The investigator is somebody who has narrowed his field down or her field down-- I think in that case, you'll have to say-- it was politically correct in those days to say his-- field down to the point where he still can-- he can't communicate with other fields. The specialization is actually disabling the progress of science. And still, he can't read all the information which you ought to be reading. So that was the problem that he started with.

But then if you go to the end of the article, you find that he's really worried about the world as a whole. He's worried about the wisdom of the human race. And being an engineer, what he has produced for the individual's problem is an individual solution. He's produced a solution, which is the size of a desk which you give to the investigator.

He has a problem. You give him this thing. And it allows him and keep track of the information.

So I'm going to look-- I'm going to spend these few minutes-- so I'm not going to talk about the web, where the web's going, or anything like that. I'm not going to talk about open architecture. It's like, oh, this is a day off.


So, I'm going to try to talk, as Bob did, without saying anything I've said before. So you've got to be prepared for a whole lot of waffle. And I'm not going to give you an overview slide either. So Ted, I think you'll approve.

I just put this thing together in a haphazard way. And I did produce some overhead slides retrospectively. And I'm not going to tell you about it. So, if you're out there on the MBone, you've probably got a link to them. So--


--pause for you to follow it. I'm going to take then two things. I'm going to be looking at the difference between the microscopic, if you like, the individual solution and the macroscopic problem of the human race. How's his microscopic solution? To what extent has it actually helped us solve the macroscopic problem?

We can take that one-- that one off. With 50 years on, we've got the mnemonics. Now we can go back. No, you can't go back. You have to you have to go back using Netscape. There's a-- well, right. A couple more. And that's one. That's where I want to be. Sorry, you'll have to look at me.

So we were 50 years on. And we're considering to what extent has it actually made things easier? Now, I should perhaps explain where I'm coming from. I'm taking the liberty as everybody else has to put in a small, parenthetical piece of retrospective. I had-- well, I suppose I still have-- I tried to write this talk to be cynical. And it's not in my nature, so forgive me if I make a mess of it.


In fact, I think we've-- I've seen this thread of bitterness, which I don't share. So maybe I'll try and turn that off a little bit. So I still have a dream that the web could be less of a television channel and more of a sea of interactive shared knowledge.

The idea is that we are immersed in something which is a warm, friendly environment made of those things which we have seen, we have heard, we believe, or we have figured out. And it's interactive.

I had met somebody just upstairs over lunch who said that a few weeks ago he discovered that, in fact, the world wide web program, the original browser/editor was in fact an editor. And, in fact, you could make links as easily as you could follow them. And that was fundamental.

It seems to me-- there's two things which seem to be totally bizarre. And one of them is the fact that you can't do that, that we've lost that. So, in fact, the thing is not interactive.

I don't know if I can't think of any hypertext experiments in the research side where you haven't been able to make links just as easier than-- the authorship has really been right up there. And now for some historical quirk, which I could go into, I have gone into, and I won't go into it, we have a whole bunch of things out there which are browsers. So that's the first thing I'm a little embarrassed about.

And the second thing I'm embarrassed about is the fact that, of course, when you made the links and you edited the text on the screen, you didn't see any of these URLs, and HTML, and all this stuff. And the weirdest thing for me, if you can imagine, is to see an advertisement and the help wanted of the Boston Globe saying they want HTML writers or HTML programmers.

I mean, give me a break. It's like asking somebody to come along with the skills to write a Microsoft Word file in binary. The whole thing is totally inappropriate. We have got past that.

So I'm assuming-- so, let's just assume for the moment, let's make a thought experiment, let's elevate ourselves to imagine that when we go back to our desktops, to the little mnemonics which we have sitting on top of our desks-- all these horrible little things-- that, in fact, that they are interactive, you can make links, you can spawn new hypertext objects to just with a command Apple splurge, whatever it is.

So now we've got that over with, the reason that that was put together was to be an interactive way of helping people work together. I wanted to be able to work with people who are off site. I wanted to be able to work with people who had a lot of enthusiasm, not very much time, and a zero travel budget.

And I suppose also these people, we were working on technical things. And maybe being introverted technical people, we were a little handicapped socially. And we would prefer, in fact, to dive for the keyboard and go make things, go type things.

And we'd like to do that in an environment where we're all making/typing the same thing. And the communication happens through that shared knowledge. That was-- that's where I'm coming from.

And I think it's likely where Doug was coming from. Although at the time, I didn't realize it. But a lot of the things he says make me feel all nice, and warm, and fuzzy.

So let's think about this problem, the team management problem. You probably all come across an organization. A couple of friends of yours, there are two of them. And they started up in a garage.

And then they were-- before you noticed it, 10 of them. And before you noticed it again, there were 25, 30 of them. And they were-- and it was great! Do you remember those days!

And it went on, and that company expanded. And there were 35. The product was really exciting.

And then there came a day when there were 50 or 60, between 50 or 60 of them. And one day, somebody walked in, and they suddenly realized they didn't know everybody there. There was somebody sitting behind the reception desk who seemed to belong there, and they didn't know who they were. And at that point, the project started to go wrong.

And you've probably seen it happen. And you've probably seen the people start to leave and start their own small companies because they gave up. And you've probably read books about it. You've seen him on the shelves in the airport bookshelves about the problem. So this is the team management problem.

And if you go back through history, history is the history of people trying to work together in teams, large or small, and making a bigger or a smaller mess of it, and with more or less disastrous consequences for mankind, great tracks of it occasionally being wiped out when people do it wrong. So this is a serious problem. And it's one of the-- and for particularly for people who are trained to look at nice, malleable, technical problems which succumb to analysis, it is a challenge.

If management of a growing team is a problem in a class, you can refer to it as management complete problems. Problems in this domain do not succumb to technical analysis. In fact, the traditional way of tackling them is with a device known as a management fad.

A fad is a heuristic algorithm which it relies on-- it has two basic properties. One is that the people who are implemented must believe in it. And the second property is that it must-- whatever it does, it must cause them to think about what they are doing. And the way it works is that because they think about it, in principle in thinking about it, they do, in fact, make a better solution to the problem they would have made had they not thought about it at all.

So here we are when looking back at history of people trying to-- trying not to, as Vannevar Bush put it, perish in the conflict before they learn a certain manner of race wisdom with our hypertext systems. If we can make itself managing team-- trademark--


--well have done pretty well. If we can go further, we should realize that the process of education, the process of politics, and all these things are things which we always-- we keep feeling we should improve on. We can set our sights quite high if we want to.

And maybe within this room, we can quietly do that amongst the converted. I'm not sure about on the net. So remove that from the record.

So, here we have mnemonics, more or less. Well, OK, so, they don't have the entire space of knowledge that is relevant to me stored in one side of the desk because we get it over the net instead to a large extent. There's a sudden change of parametrization. And the screen isn't slightly tilted, it's vertical. But basically we have mnemonics and these things glowing in the corner.

And if we haven't made good progress working in groups, then clearly we can blame the lack of-- this lack of progress on the fact that we don't have good collaborative software, we don't have good navigational tools, and the billions of people on the earth haven't all got an internet-connected computer. Is that right?

Suppose they did. Apart from the total horror of finding so many of these little mnemonics around every corner-- they'd probably drive us crazy-- would our problems be solved anywhere? Would we be-- would the race be empowered as a whole? Let's think about these scaling problems. Let's think about some large numbers.

Think about the number of documents on the web. Let's think about the number of ants in an ant heap, a la Hofstadter. Let's think about the number of neurons in the brain. I think about the number of people on the planet. What did somebody quote it as, 600 billion?

All these are numbers of things. And the interesting thing about the neurons, and the documents, and the people is that they can all have these random associations. And we're talking about random associations today because one of-- I'm talking about random associations. I'm making random associations probably. But the mnemonics is a machine which allows you to store, express, and follow random associations.

The interesting thing about all these systems with many parts is that the whole works, the whole operates because the parts interoperate. And the way the whole works and whether the whole works is defined by how the part interoperate. We can draw some analogies between these things. We can draw analogies, I suppose, between hypertext links and synapses, between hypertext links and the relationships between people. And we know that if we take that analogies too far, we have learned from experience that you don't-- that people behave in special ways, which we have great difficulty in making data objects imitate.

So we've been taught that these are basically-- we've been taught by experience. So these are separate worlds. The way that Minsky might draw analogy between the society and the mind, that the neurons and the people we treat differently. And the data objects on the web, we treat differently.

The neurons in the brain on a good day seem to work as an integrated whole. The people in the world on a good day work as several integrated holes on various different scales. The web documents just sit there.

Now, of course, pretty soon they're going to start getting up and walking around. That's going to be interesting. Pretty soon, we're going to have these things. At the moment, you've noticed the odd one you bring over, and in turns a somersault. And it's going to get worse.

So, if you think that you know what software you've put on your machine, you can expect these things to start wondering over and make themselves at home, and doing things for you on your behalf, which you may or may not approve of and understand, and bringing over their friends to help. And pretty soon you'll be-- I mean, you will be buying memory in real time.


So this is the way we're going. So the web objects are now becoming things that start talking to each other, and we have to lay down some rules. And at the moment, laying down rules for web objects is an attempt to try to get a document when it appears on one person's browser to be legible when it appears on another person's browser. When we have pieces of object code running around, this gets more serious. There are a lot-- it, in fact, it becomes quite a big deal.

I have the problem that I often have to apologize when giving talks in life. I started off the web project-- or, in fact, using a web browser, which had a style sheet by the way. We haven't even got back there yet. And I put up a-- and I would give the talks in hypertext, which meant I could take them in pretty much any order. And I got used to that.

The rules, which are the microscopic rules, if there are people-- if they're between people, we call it a constitution. We call them-- now we call them laws, we call them codes of ethics. If it's between particles in a gas, we call them that the rules of atomic interaction. If they're between web objects, we'd call them protocols.

Let's look at some associations with association as we're drawing parallels. Let's have the next one. We've had it before. Association, connection, that's what this little thing produced when you type association into it. These are associations with association, ties and link.

What associations do we have with association? What links do we have with link? This is the first association people have with link. Given a hypertext-- and for a long time, this is-- so 1989, you suggest, hey folks, put it in hypertext. The immediate reaction is, no, no, no, no, no, no, no, no. Hypertext is very, very confusing. You'll get lost in hyperspace.

And so one way or another, we managed to get the system out. And a little while later, people say, hey, hey, hey, link are cool, man. And the "click here" brigade has seized the public imagination and has been seized by the public imagination. They're rushing all over making this way cool links. And that was a phase.

And then the link now implies readership. You make a link for the guy at the end of the link. This is interesting to him.

We haven't even got Xanadu. And we haven't got trans copyright. But still, the guy is very interested in his readership.

You make the link, and a number of things happen. The guy, his ego goes up. His citation rating goes up. He starts checking the referral field, and he starts logging to find out what sort of people are reading his stuff.

And he starts to-- and pretty soon he starts writing within the document a profile of the people he's found that are-- and an explanation of how many people. And you find these little counters. Have you seen these little counters on the web that you come across saying this document is about-- and by the way, you were the 3,421st person to read it, close parentheses.

So the link comes to mean readership. And then for some people now, the association is the link with the dollar. Now that's an interesting one.

Now, it's not-- yes, it's an interesting one because it means that there's all this commercial stuff pushing the web technology. And it turns the world upside down. And it gives it a slant, which it hasn't had-- which it didn't have when hypertext was a good, old academic field. But what's interesting about it is that the link is the unit in hypertext, and the dollar is a unit in one of the simplest models of the-- of the government, the behavior of people.

So we've had two systems. We had the behavior of people, which we generally regard as something which computer science ought not to get-- computer scientists ought not to get involved in. And we had the behavior of hypertext, which, for a long time, other people felt that human beings ought not to get involved in. And we have them linked together.

And this suggests to us that, in fact, we should not simply talk about hypertext and evolve hypertext. And Douglas has mentioned already before the phrase "cool evolution" I think it was. There was a graph.

Do you remember the graph? It had the systems evolving along the horizontal axis. And it had the human systems evolving along the vertical axis.

Now, when we put something like that in, that link-- in fact, the most boring part perhaps of our society is the economic system, the marketplace. It's the simplest thing to model up to a very simple element of hypertext.

And, of course, there's a whole lot of impact, one on the other. The one would show people hypertext and global hypertext. There's an impact. There's an emotional impact. They feel empowered.

Just for those of you who are not native, who are not-- sorry, I can't say Native American-- so not an American English speaking, empowerment is a term which has crept up in the 1990s, which is to-- it means, the feeling which you feel, for example, when you first see global hypertext. But at the same time, you get misgivings, and you get fear. You get despair for normal healthy life and quite rightly so.

And, of course-- and the web is influenced by society as well. Even though we have tried to build something-- we've tried to build an abstract space out there which is full of objects. And these objects just sit in a totally inner space without anything supporting them. They are not digital fixations, as Bob was saying, of something real. They just exist. And once you've got it there in cyberspace, you can refer to it.

And what do you do? You refer to with some sort of a-- with a URL, which has got bits of domain name in it. And it's got-- and inside, it's got other things. It's got a little bit of semantics. There some reason that the name was made.

And then there's a social change, social changes that someone-- is that a company changes its name, or it splits, or it drops the project, or somebody moves from a company-- from one company to another. And the document suddenly gets another name, and we've got broken links. And so there's social change.

It's suddenly disrupted. We have a fault line through cyberspace. Because, in fact, we had a social-- there is in fact in there it is based on a social arrangement. Names are based on social contracts between the people who issue names and the people who use names. So these things, two things are really intricately connected.

When we're-- and the reason I would suggest that we haven't made tremendous advances in solving the world's problems is that we now have to do engineering of, as Doug says, the human system and the computing system at the same time. But now we're not engineering mnemonics. Now what we're doing is we're-- in fact, we're engineering the greater machine, which is made out of all the links.

We're constantly changing these microscopic rules. Physicists are interested in the way the microscopic rule-- if you invent microscopic rules, you can then deduce the macroscopic behavior. And an engineer creates microscopic rules and creates macroscopic behavior.

For people, governments are changing the laws. They're tweaking them here and tweaking them there and hoping that people will follow them. On the web, organizations like the World Wide Web Consortium are introducing new aspects to the protocols and hoping that people will follow those protocol specifications.

And these protocols, who cares about protocols? Nobody cares about the protocols. But they care about how the whole system behaves.

They care whether when everybody tries to-- if we put up the URL of a slide, 100 people watching, and they all go and access the slide, and the server can't handle 100 hits at the same time. Or if it's a shoemaker [INAUDIBLE] commentating you do better, they can't-- it can't handle 100,000 people hitting the same place at the same time. They're interested in the fact it doesn't work. They're not interested in the protocols.

And then we change the protocols underneath and so the objects can migrate about through to the network and be replicated. We've changed the overall macroscopic behavior. We can change its tolerance defaults. We can change its availability and its response time, the way it scales.

When we're looking at the goals which we're aiming for, the interesting thing is that for society, the goals are often expressed in macroscopic terms. A goal of society is that it should be fair, that any two individuals should have a fairness relationship. There are rights to-- rights to free speech, to health, clean water.

And we also have constraints on the whole. We have constraints on the whole of society, that it should be stable. It shouldn't wipe itself out. We shouldn't have-- the safety of the race is important to us. And we have to try and satisfy both of those.

Whereas for the web, in general, the little protocols don't affect us. But we are also interested in things like heterogeneity, the fact that the protocols will work independent of platform. These are microscopic rules, as well as the macroscopic rules, that the whole thing should work if any one link should be cut.

When we do this engineering, what we're doing is we're creating a different topology. We're playing with the topology. We have our mnemonics now. We can make links.

What's the topology of the result? When we make the big machine out there, what's the topology? What's its quality of our society? What's the topology of our hypertext going to be?

And as we design things in engineering, we come across a number of interesting things, interesting, useful properties the way that-- well, in the internet world, for example, the internet world, anybody that's come through there is totally sold on decentralization. The internet works because it's decentralized. You can start an internet network anywhere. And then you can connect it up to the big network without too much of a horrible frightful change when you do that.

And the same with the web, you start a web server anywhere. You can start a little web. And then when you connect it, [INAUDIBLE] it up to the new one, there's only an incremental change.

And internet people have come through this. If you look at the Internet Engineering Task Force, you'll find a whole bunch of people who've come through with a pretty strong political, anarchic, decentralized view as well. They just decided to take this from one system-- the philosophy from one system and apply it to the politics as well.

And you find that, for example, the newsgroups on the internet have no home. They just exist in the network. You can't pin one down. A newsgroup starts somewhere, I guess, but-- and so do the articles.

And they propagate, and the network sees to their propagation. And everybody ends up with more or less the same view. But there's no one particular home. So that is something which I think just an example of one general rule that we pick up.

But there's another one. I could give a number of examples of the rule. I could give a number of examples of decisions that we're having to make right now as we design the protocols, which could seriously affect the topology of the web. And that could effectively seriously affect the topology of society.

And an interesting-- that this decentralization is, in fact, often a myth in fact, I should say in parentheses. Because very often you have a nice decentralized system, and it all works courtesy of-- at another level, there is a centralized, a totally centralized system which is policing it. So the web is totally decentralized, but it does use DNS space. And DNS, the domain name system, address base, this, is, in fact, a singly-rooted tree.

Now, in fact, the web doesn't make any assumptions about that. So in fact, the web behaves in totally decentralized fashion. But it does rely underneath on a centralized system.

And when you see people happily go into the market, and bartering, and chatting, and talking to each other, and exchanging goods for money, and what have you in the marketplace on a Saturday morning, the reason that it all works in such a nice fashion is that there's a policeman standing on the corner by the statue. And behind him, there's a hierarchical system going up to the Supreme Court in this country which guarantees that. But when you're talking about the marketplace, we don't discover-- we don't discuss that.

But then there are other typologies which do scale. Well, when you look at the-- when you take off the hood and then look underneath, they have the same properties. They're basically fractal. Remember fractals? They were popular 10 years ago.

This-- next slide-- I went looking for a fractal picture. I wanted something which illustrated something which was basically trac-- which basically had scale and basically it didn't. I went two clicks from my home page. And you see, we got two things.

The interesting thing at the bottom are the trees. We like trees. We like trees because they've-- even though they're not in color, they have a shape and they have a tree structure.

Now, I'm not saying that-- fractals, of course, the most interesting ones are not trees. This is rather a simple example. But then as you zoom in, you find that there is an interesting structure at every level.

By comparison, look at the building. We have dominantly at the scale of about 100 meters, we have a rectangle. As we zoom in, we see nothing until we get to the one meter level, which we see another rectangle.

And then we can zoom in. And probably until we get to the rivet level, we won't see any more detail. So this is basically in a nutshell the problem with modern architecture. It's not fractal.

And if you apply it to protocols-- it's very interesting to make protocols which are fractal. If we can find a design which scales in the-- you can, for example, make a structure in your hypertext which works at a lower level and use the same sort of structure at a higher level where you'll lose the detail, then that is very interesting. It has these interesting properties. It had good scaling properties.

Think about society. Does society have these scaling properties? Yes. It has structure on lots of levels.

It has an international structure. It has national structure. It has state structure. It has structure by field. People are structured into guilds, and groups, and families, and people.

And when you look at the microscopic effect, what's it like to actually be in a fractal environment? Because after all, if we're trying to make a fractal environment for hypertext, we can't decree that it is. We can't say, you should make a-- we can't design at a top level. We have to only make the protocols.

We have to design, for example, sets of link types. Or we have to put in transclusion or not put in transclusion. We have to make little decisions at the microscopic level and hope or demonstrate that that will lead to something with interesting topological properties.

So, let's think about society. What's it feel like to be in a factual situation? It's essential. Every human being needs to be in a fractal environment.

Every human being needs to be part of a structure at a small level, at the family level, at the next level out, and to have some feeling for where things are going on a global scalar. Everybody needs to have some links, lots of local ones, lots of not-so-local ones, not some very not-so-far distant.

When you're putting to-- when you're managing a big team, an organization, planning those things so that there are communication paths of different lengths is very important. I call it "information plumbing." Sometimes it's really, really crucial when you've got-- when you're trying-- the consortium team is a good example.

We've got all of research on one side and all of industry on another. And we're trying to allow good ideas to filter through. And we've got some key people we can recruit in various places. And we have to put in the plumbing, decide, OK, that this person is going to talk to this many people at this level, this many people at that level. It's information plumbing. And when it's done well, people feel comfortable.

That's interesting, isn't it? That a human being feels comfortable when they have interactions at a wide-- at widely diverse scales. Maybe human beings are programmed in some way such that when you put them together, they will interact so as to form something which has got this fractal properties.

Maybe their program-- maybe this comes about through evolution because animals that didn't work like that died out. They couldn't make a big society work. They couldn't make the society scale.

I noticed Bob made a remark about the problem that no one person knows everything in the system. One of the interesting things about the fractal property is that you can move from the small scale-- from the specialized area, you can get out. You can get into bigger-- and without following too many links, following a finite number of links, you can get back to anywhere, so that we're looking for a system. One of the properties of this system that we need is if there's a problem and a solution, there's going to be finite number of links between them.

So, I think I'm coming to the end of my time. And can go on, take that appalling building off? So what are we heading for? We're heading for making local decisions which have global effect. We're making decisions all the time that I can throw some at you.

What about end-to-end confidentiality? Is it reasonable for two agents, whether they be objects or people, no matter where they are geographically located to be able to communicate without other agents, be they objects or people, being aware of the communication? That's a nice little rule we can lay down.

It can determine the topology of society. It can determine whether you can run a police state. It can determine whether terrorists can, in fact, set up shop and pull down society. And this is up for-- and this is up for grabs now. It's up it's being discussed by Congress, and technologists, and lawyers all over the place.

Should we put in-- should we assume the trans copyright model? What would that do? Suppose we do that? What would be the global effect? Should we go for something much more complex, which would have lawyers involved very, very much more?


What about the right to anonymity? All these little things, the little microscopic rules we could set up, we set up the microscopic rules. Anonymity, confidentiality, we could do that with a protocol. We don't have to do it with a law. We can implement the protocol, and then we have a look at society.

So these are real-- these are rules that we have. We have rules existing about the web at the moment. We have a rule, for example, that-- there's a basic rule that defines the usefulness of a URL. That if I dereference a URL, and I get something, and I'm interested in the thing I get, and I give you the URL, that's got to be as good as giving you the object. When you dereference it, you've got to get-- in some way, even if you only get a black and white copy, you've got to get the same thing. So we have a rule about that.

We have a rule that when you get something, that may not have side effects. With these rules, you can start deducing things about the system, and you can start making more sophisticated systems. It can allow you to do caching and this sort of thing.

When we go into the world of mobile code, of secure systems, network payment, these principles are being laid down, sometimes silently, and sometimes amid uproar. They're going to define the behavior of the new machine, of the big machine, of the new ant hill, of the new brain, whatever it is, which is the sum of ourselves and our data. Can we have the next slide?

And always, the question is, can the applications of science-- can they allow the investigator to encompass the great-- the great record and to grow in race experience? Can the whole human race grow out of this? Or is it just a question of techies having fun? Thank you for listening.


MODERATOR: Thank you. Thanks so much, Tim. We now have some time to probe a little deeper. Anybody wants to ask, please go to the mic can identify yourself. And don't forget, we have a panel session at the end of the afternoon if you don't get satisfied just now. Michael.

AUDIENCE: Michael Lesk. You talked about the question of whether we want anonymity as a technical one. And the problem is that, in fact, this is one where the legal process is moving very strongly. We've got a variety of--

BERNERS-LEE: Sorry, I'm being directed.

AUDIENCE: We've got a variety of people in politics who-- somebody referred it once as "the four horsemen of the net--" child pornographers, drug traffickers, foreign agents, and terrorists. And the question is whether--

BERNERS-LEE: Can we add to that list?

AUDIENCE: By all means. But one of the questions is, we have to think if we decide that there will not be anonymity on the net, will we have a better chance of staving off some worse thing that the politicians might do to us or not? And I happen to feel that at the moment that's probably the best thing to do in the realities of the situation. But I don't really see the right debate technically among people over how many of these things and we do, and what should we do?

BERNERS-LEE: I think-- my thesis today is that you can't distinguish between the technical question and the sociological question. So the anonymity-- you can put anonymity in simply by setting up a proxy server with the required characteristics. And it's been done. And you've made a sociological decision by doing that, at least by throwing away the logs.

And so we could go into these various-- I think that it will be a rat hole if we go into them too deeply. Are you asking me to say whether I believe in anonymity or not?

AUDIENCE: No, I am saying that you could use-- everyone could establish an ethical [INAUDIBLE].

BERNERS-LEE: Sorry. So say it again. You repeat it for me, and I'll repeat it for them. [INAUDIBLE] I didn't catch.

AUDIENCE: What I'm saying here is if the technologists get together and say, well, yes, someone can build an anonymous server, but we'll refuse to exchange traffic with it, we can probably stamp it out. I mean, there are lots of technologies that get stamped out by legal means. We're pretty far along at the moment to abolishing nuclear power, childhood vaccination, and birth control research as a combination of legal and political aims. I'm worried that the internet might actually be more endangered than we think.

BERNERS-LEE: Yeah, OK. Well, my first answer to that is it is a very good-- it is very appropriate for technologists to produce protocols which allow people to behave reasonably. So, for example, if there is pornography on the net, which is quite a good example, at the moment, there is no way of declaring that, of covering yourself, and allowing the server to say to the client, hey, this stuff is bad stuff. I want you to take full responsibility for the fact that you are over 18, and you're not going to show it to anybody else before you sign up.

If you put in some protocols which allow people to behave reasonably, and to flag that, and to behave more as they do in real life, then that probably-- that may stave off a backlash, people trying to stop-- stamp on the whole thing because it's just unreasonable. And the backlash would be totally reasonable. If you make protocols, which don't allow people to behave reasonably, then you're creating an unreasonable system.

And in this particular example, for example, the W3C has an initiative called PICS, Platform for Internet Content Selection, which is doing exactly that, it's putting in technology for allowing people to label things, like third parties to label it, to label things, allowing teachers to set up filters, allowing everybody to protect children, and incidentally being a lever for allowing labeling metadata of all sorts of interesting types.

And that will make the internet, the web, a whole lot nicer place. And it may be that when it comes to anonymity, the same sort of thing.

My personal feelings that probably when you browse the web, what I would like to be able to is browse a number of modes with a number of different personae so that an eyebrow-- either browse anonymously, or I can browse as me with my name, or I can browse-- if I have to browse through a catalog, for example, they will-- or if I browse a magazine, which has got advertising in it, the magazine will probably not want to know my name. But it will want to know how old I am, what salary range I have, how many children I have, my shoe size, and whether I have a backyard, and this sort of thing.

So I may want to build a switch between those persona. And we would need to lay down rules which prevent people-- principles which say that you are not allowed to tie these two things together, even if you do manage to-- even if you write programs, which allow you to piece these two things together, the codes of behavior which say that if I give you this information, I'll give it to you on the following terms.

So in the same area, we can lay down some reasonable ways of operating. We can define protocols which allow exactly the sort of negotiation you mentioned whereby you may be browsing anonymously, and you get to somewhere where it says, sorry, we don't take anonymous people. If you want to come into this door, you take the hood off, sir.

And that's perfectly reasonable. That's reasonable behavior. We need to allow it. And so we made protocols for it. Paul.

AUDIENCE: I want to ask you-- Paul Kahn. I wanted to ask you a question about your comments on topology because it's recently come up in our work. And it was also one of those things that in some of the hypertext systems previous to the world wide web was of great concern, and then everyone was amused that it didn't really exist in the world wide web, and that a link-- an HREF is an HREF. A URL's a URL. It doesn't really matter where it is.

If you want to make a link from one thing to another, you just-- the nice thing is it's simple. The not so nice thing is that it may be the fastest way from Singapore to Hong Kong to Australia back to MIT. But as you move from one place to another, you often have no sense of the topology of the structure of the content of the material. There was nothing in the way in which things were coded-- or precious little in the way in which things were coded that gave you a sense of where anything was in relation to anything else. And I was wondering if you're suggesting that there needs to be more of that built into the coding, or you're just suggesting that we'll discover that it's there?

BERNERS-LEE: Oh, you bet. I mean, we've got to take a lot further than that, too. I have to confess that 10 years before writing the world wide web, I had a program which I developed personally for my own use which had links, and it had typed links.

And that-- and originally, the web-- there was an HREF, and there was a relationship equals. And there still is in some of the specs. There's a rel equals. And in fact, in some cases, you see it used as the relationship between the document and its author.

But the moment you're allowed to put typed links in, you can do two things with them. One of them, which addresses, I think, your question is, you can make structure. So you can make relationships. You can say this is a table of contents, this is-- the links in this are basically I intend as a trail so that you can bring up some browser gadget to allow somebody to go backwards and forwards through these links.

This the table of contents. This is a summary of that. You can make relationships between the documents.

And the other interesting thing you can do, which is maybe more interesting, is if you regard documents as being representative of things, people, concepts, if they are in fact information about something, even if it's another document, but about any arbitrary concept, then when you make relationships between these things, you can make relationships-- state relationship between the objects you're talking about. That becomes very interesting.

Then we have things like [? Aquanet ?] in all the hypertext systems, which have had-- which allow you to express argument, for example, which allow you to express a legal reasoning where there is semantics in. And I think that perhaps people have abandoned semantics because they felt that, hey, we can't sell the AI problems, so let's give up.

But in fact, to put some semantics into the web would enable so many things. There are so many applications which become trivial. When you invent a few linked types, you can represent the whole thing as a bunch of linked objects.

It would enable computers to help you because the computers could read the linked type. In the moment, reading, as you say-- I've got a bunch of slides on this. I made them 14 months ago. And I give them every now and again.

If we use that rel field, then instead of it just being a mass of objects with these anonymous links between them, then the links actually start to take-- the links have significance, and you ask questions. You can ask interesting questions like, OK, so if this guy dies, what are all the things which he's written which have not been-- which have not got-- which are not described by documents which he has also written, which are used by things which are recursively transitive closure of used by things which are used by anything which I am relying on, and to do interesting traces like that, to ask logical things of the network.

So that, then, I think will make the web very much more powerful. It'll allow people to put information on which has got-- which have got semantics, which computers can then use to actually help them. And in all this, it's a great system for communication between people. But the computers are all part of the infrastructure.

And some people like to think that one day the computers will actually start to participate, too. Other people don't. And they have given them up. But some people-- there are a few computer scientists now and people back out there who think that Monday, the computers should have a presence on the net, too.


AUDIENCE: Ron Becker, University of Toronto. I want to go back to your original goal of facilitating scientific research, which is very much in the theme of the tribute to Vannevar Bush. Could you describe what you regard as the one or two major dramatic successes in the use of the web to facilitate scientific research, or describe what you think are one or two major limitations that you think would have to be overcome in order to attend to your original goal?

BERNERS-LEE: Well, I'm conscious as ever, of course, of the limitations. The big limitation is the fact that the thing is not interactive. So the original idea that we could share knowledge, and annotate each other's, and make caustic hypertext comments on each other's documents, which is really interesting in scientific environment, for example, in paper review or in anything, in back of an envelope writing, is just not there. The web has got this strange publishing model that is in a sense trivial to fix it-- in a sense. So for the purpose of this talk, I've assumed that it's fixed.



AUDIENCE: In what sense [INAUDIBLE]?

BERNERS-LEE: In the sense we know exactly how to do it, and we've done it. It's been done trivially. It's trivial in the sense that we have to try to persuade people who actually make the leading products that they should give this to people. They should replace the hot list metaphor with something which is that if they want hot lists, you should have hot list objects on the web. And if you want to have hypertext objects, you should have-- you should be able to edit hypertext objects, that basically anything which you can read, see you ought to be able pull apart and manipulate, then that is difficult to get that across. It's difficult to get it across because the commercial push is from the people who are disseminating information, who are in output mode.

But I have a lot of hope because I know that the online services, for example, they all started off saying, hey, what we're going to do is sell information. You'll dial up with your telephone and your modem, and we'll sell you whatever information you'd like. And what they found was they want-- the people wanted to talk to each other.

So people really like to type. And they like to input, and they like to be part of it. And I think that pretty soon the moment the first really neat hypertext editor comes along that's integrated happily with a workspace and all these other funny things on the screen there, I think it's going to be-- it's going to be a walkover. It's going to wipe the floor with the competition.

MODERATOR: We're about to fall off the edge of the tape soon.

BERNERS-LEE: I'm about to fall off the edge of the tape.


MODERATOR: --and we continue with this line of questions during [INAUDIBLE]. OK, so, the next thing up-- thanks very much, Tim.