Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS

 

Main NerdTV Page NerdTV Archives
 
send feedback rss/podcasts I, Cringely Sponsored by Datapipe Creative Commons License
transcript

NerdTV #12: Bob Kahn

Bob Cringely: So, welcome to NerdTV.

Bob Kahn: Thank you.

Bob Cringely: It's really great to talk to you again. It's been seven years. The last time we spoke, of course, it was for my Nerds 2.01 Brief History of the Internet and now we're talking for NerdTV, and it's more about you. How did you get interested in becoming an engineer or a scientist?

Bob Kahn: I'm not sure I know how to answer that question. You know, when I went into college in the first place I guess my thought was I was interested in building things. I was interested in how things work. I was sort of interested in mathematics. I was interested in science in general. I wasn't too skilled in the arts. I never saw myself as a writer or an actor, certainly not somebody in politics. I just didn't see myself in any of those fields, so I thought I'd try industrial engineering, but there really was no program that I could get into where I was going to school. And so then I said, "Well maybe chemical engineering," but I didn't like the lab work, so I ended up in electrical engineering. And the thing that appealed to me about it was it kind of had a very rigorous mathematical basis to some of the areas of that field, and you could actually build your own things. So I got into electrical engineering -

Bob Cringely: And where was it?

Bob Kahn: This was at City College in New York. Actually, it was a Queens College/City College program, and when I graduated with that degree I went to work at Bell Laboratories, briefly thought that little practical experience would be a good thing before going on to graduate school. So I went to work at the original headquarters, or at least it was headquarters of Bell Labs. In those days it was at 463 West Street in New York, right on the Hudson River, so you could actually watch the Queen Elizabeth and Queen Mary coming in every other week and coming in and going out.

Bob Cringely: What year was this?

Bob Kahn: I started at Bell Labs right after graduation, which was, I guess, around January 1960. And I was at the labs for a number of years, although I went to graduate school in September of that same year, so formally I was there full time for maybe nine months, and then I was back there summers for a number of years after that. That was before they had built Homedale in New Jersey, and so that was really where the headquarters was, but much of the technical work took place out at Murray Hill, New Jersey, which the last I saw was a Lucent operation, and Whippany, New Jersey. They also built a plant out at Homedale maybe a few years after I joined them, and eventually they got rid of the New York plant completely. I think it became a condominium of some sort. Building is still there.

Bob Cringely: Was it fun?

Bob Kahn: Oh, I loved it. Bell Labs had a certain culture and rigor to it, and it really was a national resource back in its day. Of course, it's now gone. But it was a different time - a different age - and it was kind of a great experience for me. I worked with some wonderful people there. Got to learn about big systems, and unlike many of my colleagues who went into the more hardcore technical pieces of the business like color television or coding theory or various development of magnetic core devices, things like that, I got into more the systems aspects of the Bell systems, things like: How do you place lines in the Bell system for efficient performance? Where should they go? How do you maintain the cost effectiveness of the system? How do you guarantee certain quality of service and degrees of performance? And so it was very much of a mathematically-oriented analysis of the very complex system.

Then I went to graduate school. I got a NSF fellowship and went to Princeton that Fall.

Bob Cringely: It seems to me that this really prepared you for the kind of work that you were doing later on.

Bob Kahn: You could say that now, but certainly I had no idea -

Bob Cringely: You had no idea?

Bob Kahn: No idea back then. So it was just a really interesting job, worked with some wonderful people, very motivated, very dedicated, very smart people, and then I ended up going into graduate school. It was a small class. Princeton was really just building its department back then.

Bob Cringely: And this was still electrical engineering?

Bob Kahn: It was still electrical engineering, but most of the work that I did was really math and physics kind of work, a lot of the self-study over there, and they gave you ample opportunity to figure out how to fend for yourself. So you could take classes - you could do self-study. You had a variety of requirements that you had to meet, and of course, tests that you had to pass, and then, of course, you had to write a thesis. And so, I finished the thesis in 1964 and then ended up becoming a faculty member up at MIT right after that.

It's interesting. I actually got an honorary degree from Princeton in 1998, even though I have a regular doctoral degree, and at first I thought they made a mistake and they had just forgotten or something like that, but they said, "No, no. This was intentional. We really intended it." When my thesis advisor retired from Princeton, most of his students came to wish him well and give him a nice send off, and when he came to introduce me I was like student number six, and my advisor was a gentleman named John Thomas, a very distinguished gentleman who now lives out in California. When he got to introduce me he said, "Well, Bob was student number six, and he's the only student who ever wrote two Ph.D. theses," and I'm scratching my head trying to figure out how come I don't know that.

Bob Cringely: (Laughter)

Bob Kahn: Well, what happened was he had done enough work early on that we knew it was sufficient, but we hadn't spent enough time in the program, so we assigned him to work with another faculty member who is actually still on the faculty and a good friend of mine. He and I worked together and we did some additional work. And so the thesis was sort of a patchwork of an A plus B thesis, and so when I got up to -

Bob Cringely: They manipulated you.

Bob Kahn: No, no, no, no. It was great time.

: I enjoyed every minute of it. But when I got up to Princeton and they said, "Could you say some words about your time at Princeton?" that was one of the stories I told. I said, "You know, quite frankly, I thought they had made a mistake, but when I sort of thought about it I realized since I had written two Ph.D. theses, according to my advisor, it was really only fitting."

Bob Cringely: And effectively that was your third.

Bob Kahn: No. Two theses, one degree.

Bob Cringely: Oh, one degree. Oh.

Bob Kahn: One degree.

Bob Cringely: So you went to -

Bob Kahn: And that's also the time I, you know, in between finishing all the course work and the studying and writing the thesis I spent quite a bit of trying to hone my golf game that summer. I just decided to back away for a while. I explained how the other thing that we had done that year got our golf scores really in good shape by playing four rounds a day, seven days a week, for something like four months or five months in a row.

Bob Cringely: That's 72 holes of golf a day?

Bob Kahn: Walking, carrying our bags. And I played with another fellow there who was a really fun golfer to play with, and it turned out he was actually at the meeting, so I had no way to verify the story. It was actually at the meeting where he stood up in the back where he used to sit and said, "Hi Bob," and it turned out he was the Chairman of the Physics Department.

Bob Cringely: (Laughter)

Bob Kahn: Small world.

Bob Cringely: Small world. And you still golf?

Bob Kahn: I do, but not very much, and I don't play very well. The older I get it seems to me the less distance I get and the less - but to be really good you gotta play a lot. Because it's all in muscle memory. It's all how you feel out there. It's like piano playing. If you gotta think about what finger to hit what key next you'll never play well. Fingers learn how to hit the keys, and the brain is just forcing the fingers to do what they know how to do.

Bob Cringely: That's right.

Bob Kahn: So if you're playing all the time, that's sort of the first step in getting good. Oh, you have to have some talent, as well, but if you don't play all the time you'll never score really well. At least in my opinion.

Bob Cringely: So you went from Princeton to MIT. Can you characterize the two institutions?

Bob Kahn: Well, they're both well-known institutions, obviously.

Bob Cringely: How are they different?

Bob Kahn: Well, Princeton was a rural school. We were an hour from New York City. I used to go in periodically, take the bus 'cause I had gotten subscription tickets tot he Metropolitan Opera, so I used to go to watch opera performances every now and then, which was about the closest you could get to big city culture was to take a bus ride downtown. They had quite a bit of local activities at Princeton, but it was mainly a rural kind of culture. Wonderful learning experience, a lot of social activities. Everything that goes along with a college town.

Football during the Fall weekends and basketball during the year and so forth. MIT, of course, is right in the middle of a large city. It's in the Boston area and Cambridge, right across the Charles River. Very intense, very motivated people up there. People who are really trying to make a difference in the world, either in terms of inventing something or making something better. Very intensive, almost compulsive environment. Stay on top of your field. And I think one of the things that was true for me at MIT when I was there - I made some very good friends, a lot of lasting friends, and I think every single one of them was intensely motivated to make a difference. It's something that you don't get at every school. Many people think an educational experience is an educational experience just so you can learn and be able to practice in your field, and here the motivation was much more around making a substantive difference. I remember I was sort of the mathematician on the staff, although they have some very good mathematical people, including Claude Shannon, who was mathematician from Bell Laboratories, well known for his work in sorting information theory, but an awful lot of really good people. I knew I was good, but the thing I was lacking was any practical experience. I had come up through a much more theoretical kind of a background so the gentleman who ran the group and one of the people whom I really looked up to, a fellow named Jack Wosencraft, took me aside one day when we were just having kind of a friendly chat about how things were going, and he said - I don't know whether I asked him whether he had any advice or whether it was just gratuitous, but he said, "You know, the thing you might want to consider is just taking a year or so off - take a leave of absence - and go apprentice yourself to somebody who knows how to build real things." So he said the thing that's different between yourself and all the other members of the faculty - they've all been involved building real things, either at Lincoln Laboratory or in the Army Civil Corp, or in industry, and I had not had that experience. So, I really trusted this fellow's judgment.

Bob Cringely: That's what led you to BB&N?

Bob Kahn: And I decided virtually on the spot to take a leave of absence, and he gave me a list of names of people that I could go talk to. I ended up deciding to go to Bolt, Beranek and Newman, which at that time was a small consulting firm in the Cambridge area. It was maybe about as far in the other direction from where I lived as MIT was in the other direction, and I met a gentleman there named Jerry Elkind, who - and Jordan Baruch, actually, was the first one I talked to there. They persuaded me that that was a reasonably good place for me to consider, and so I took a position there and I started to work on computer networking, thinking that that was a really interesting area to get your fingers dirty in. I wasn't thinking about building networks at the time. I was thinking about doing something more practical. Rather than solving math problems, figuring out how to design a system that might actually perform a function.

And that was what I was working on. I had not heard of DARPA at the time. It was called ARPA back then, although BB&N had some support from DARPA, or they had just gotten some support to work in the computer field. I didn't consider myself a computer person at the time. My background was really more communications, and as a result that's what I thought I was working on, a communications problem. But an awful lot of people around me in that very group who were at the forefront of computer research, and I got myself totally immersed into the computer side of things, as well, people working in operating systems and languages, and people who I still hear from to this very day. I think that was an environment that made a big difference for me because it suddenly gave me a visibility into the computer side of things that augmented what I was doing on the communications side - motivated me to think about networking for computers, rather than networking just for the point of moving bits around. In fact, if you looked at the history of information theory, they almost distinguished themselves in that field from not wanting to apply any semantics to the bits that were being sent.

Bob Cringely: That's right.

Bob Kahn: And here you were talking bout how would the computer use it and what's it for and things like that. It was rather a different take on the subject.

Bob Cringely: So in essence a bit isn't a bit?

Bob Kahn: A bit is something for the computer to use.

Bob Cringely: Well, I know, but I mean from the information theory standpoint a bit is a bit is a bit, but if you need the bit to arrive at a certain point to perform a certain function in a computer program then that bit is more important.

Bob Kahn: But I wasn't actually thinking at that level. It was more like how do you get a computer connected to this thing and what would a computer do with it in terms of how operating systems work, and the like. So it was a pretty interesting experience for me.

Bob Cringely: Now this was before the ARPANET?

Bob Kahn: Yeah. This was before the ARPANET. This was roughly in 1966 by then.

So I had been working on this and a lot of people were telling me that this was just sort of a waste of time because networking wasn't really likely to be a very important field going forward, and they could have been right. I mean, I didn't know, but my gut feeling told me that that wasn't right - that computers really had a more important role to play. Those were the days when batch processing was really new de riguer. That's what most of the big companies had, and there were only probably a few hundred companies in the county that really could afford the biggest machines. So we weren't talking about individuals having personal computers and any of that sort of stuff. And time sharing hadn't really taken off in a bit time. I mean, people had built some early timesharing systems, but they were very specialized, mainly in the research community, and yet the networks really were intended to deal with the interactive timesharing systems. This was the beauty of the vision that J.C.R. Licklider had in the early days when he talked about interactive computing and man-machine interaction and networks to cause them to all be linked. So these ideas were swirling in the community, but people really hadn't focused on how would you actually build something like this to actually make it work. And I started to work on the details: How to actually design such a net? How would you make it work? Not thinking that we would actually build it, but we'd just go through the exercise and then I'd end up back in teaching.

Bob Cringely: Did you -

Bob Kahn: But along the way I learned about DARPA.

And not only did I learn about them, but I learned that they were actually interested in causing a network to be built and I was asked to sort of share with them some of the memos I had been writing on different issues, whether it be error control or flow control, or you know, whatever, and DARPA ended up putting out a bid, and it was called an RFQ, or request for quotations, on a network that they wanted built. I ended up writing the technical part of the proposal for BB&N. I didn't even assume at that point that I'd be involved in building it. I figured I'd just help write the proposal, but it turned out that we won the proposal and that we needed somebody to be kind of responsible for evolving the system design as time went on. So I decided to continue on at BB&N to see the network actually get built, and of course, once we had done that I got offered a job by DARPA to come down and eventually ended up running the whole office down there. I do white water canoeing as one of the other things I like to do, and so this was like putting a canoe into a raging rapid, you know, and once you're in the boat it's hard to get out. You've gotta sort of - - shoot the course.

Bob Cringely: So you thought you were writing this proposal that might go nowhere?

Bob Kahn: Well, no. I knew it was going to DARPA.

Bob Cringely: I know, but it might not be accepted.

Bob Kahn: Any proposal can be not funded.

Bob Cringely: That's right. And instead it changed your life.

Bob Kahn: It did. I was involved with a set of other people at BB&N in a group that was led by Frank Hart and involved some other people there that were really kind of system developers. These were people who knew how to build things, and I sort of came on with mainly the ideas, and so together we actually ended up putting together a system and actually built it and deployed it. There have been books written about that. There's a very interesting book that Katie Hafner called Where Wizards Stay Up Late. Katie and I talked a lot about what was in the book 'cause in some of her early drafts I felt she didn't understand the full story of what had actually happened. Boy, it's a really good read and for the most part it tells the story in an interesting fashion. Mitch Waldrup wrote another book called The Dream Machine, again a similar kind of thing from the point of view of the computing industry, building off Licklider's points of view. So I was involved in that right to the very end. Once we had it to the point where we knew it worked and had been deployed, I ended up putting on a demonstration of the net, the first public demonstration of the network, at the Washington Hilton in 1972. It was October of 1972. We got contributions, a lot of different terminals. It was the event that caused the network to become real. We had to get the machines working. We had to get the protocols working. We had to get everything up to speed.

Bob Cringely: It was the deadline.

Bob Kahn: It was the deadline. Right. It was the final exam, and right after that finished I became a DARPA employee and stayed at DARPA for about 13 years.

Bob Cringely: Now, at that point in '72 you had started in '66, '67?

Bob Kahn: With BB&N. Yeah.

Bob Cringely: And so you had multiple nodes on the network. I mean the first IMP was at UCLA.

Bob Kahn: The contract was awarded in I want to say January of 1969, and it was kind of a nine month delivery cycle for the first node on the net. We called it an Interface Message Processors, or IMPs. These were really packet switches built out of mini-computers. We used Honeywell 516s.

Bob Cringely: Yeah. The size of a phone booth.

Bob Kahn: Yeah, or a refrigerator really. And heavy as can be, but they worked. They worked out the box, so to speak.

And nine months we had the first one, ten months we had the second one, eleven months the third one. So by December of 1969 we had a little four-node network deployed on the West Coast. One at UCLA, one at Stanford Research Institute up in Menlo Park, one at the University of California at Santa Barbara. That was the third one. The fourth one was at the University of Utah. So a triangle in California and a little spur out to Utah, and based on how well that worked DARPA was reserving the right to put more nodes on the net, and of course it worked so well that they immediately went into the second phase and that turned it from a 4-node net into a 19-node net and then eventually it became over 100 nodes until it was finally split into pieces in 1983. One piece remained a kind of research net for the research community and another one that was taken over by the military for some of its more operational kinds of needs.

So they redeployed those nodes to military bases they were comfortable with rather than leave all the nodes on let's say university campuses. So that was really the very first computer network that was every built. We were just very fortunate to have been involved in it. There were several people who were involved, as well as myself. There was a gentleman at DARPA named Larry Roberts who managed the project. Larry was a kind of - I think of him more as like the producer or an architect. He was sort of masterminding the effort, but as I later found out 'cause I had the job that he had shortly thereafter. You can't really become an implementer from that point of view, and so Larry was really orchestrating it, and I was doing the hard core systems design, and this other group at BB&N were really responsible for the implementation. They would say, "Get the hardware built. Get it checked out. Write the software. Get the protocols done." And in addition there were a set of people in the community who were involved in getting their machines connected and helping to figure out what protocols the computers would use to talk to themselves because it's one thing to move the bits around. It's another thing for the computers to actually make use of it. And that's the basis on which I ended up meeting Vint Cerf, who I then collaborated with on the Internet side of the story, which is really something that started after I got to DARPA.

Bob Cringely: So you're at DARPA and having participated in building the ARPANET and specifying it and doing a lot of theoretical work. It had protocols, but when you were at DARPA you felt the need to develop a new protocol. Why did that come about?

Bob Kahn: Well, there was quite a bit of theoretical work on the ARPANET, but it was mainly done by people like Len Kleinrock out at UCLA when it had a number of students that were looking at ways of predicting performance and analyzing things, and there was quite a bit of work done at the simulation level by Howard Frank and his people at Network Analysis Corporation up in Glen Cove, New York. We were really doing the engineering of the system. We were building it. There was not a lot of theoretical work going on. We had to design algorithms; we had to figure out how to make things work, but we ended up with a network that allowed you to connect a computer to it. The computer could say which of the output wires essentially on that network you wanted the bits to go to, and when you put them in it would say where they were going and when they popped out it would say where they came from, but it was all contained within one net. Now if you wanted to have another network involved there was no way within the constructs of the way the ARPANET worked with the host protocols worked that would let you identify a random machine at some other part of the world on some other network.

Bob Cringely: No way to make an INTERnet.

Bob Kahn: Right. Not with the effort to recreate it.

Furthermore, the ARPANET had a number of very important features that were really useful for computing, namely, it was intended to get the data reliably to the other end - just directly, you know, so if the data didn't get there the basic assumption was something was broken. So if you were going to print something on your printer and it didn't print you would assume there was something broken there; whereas, if you're in a more realistic communications environment - let's say with radio communications - you may not be communicating for some very good reasons that have nothing to do with something being broken. Like if you're in a tunnel, a radio wave is not likely to get to you unless they've taken steps to radiate the signal in the tunnel. If you're behind a mountain the signal may not get through the mountain. You could be in an area where there's a lot of interference, intentional or otherwise. So there are many reasons why in - let's say a radio net - you just might not be able to communicate, but nothing's broken in the canonical sense of what that term means.

And so I knew we needed to have a different way of dealing with it. We had to deal with the issue of unreliability along the way, and we had to deal with the issues of addressability, and those were two of the main issues that led to the re-conceptualization of the protocols and we called it TCP at the time and later we took the part of it that dealt with the Internet protocol and sort of broke it out and made a separate thing. That's where the name TCP/IP came from. So the IP part had to do with addressing the machines and having the networks route based on IP addresses, and the TCP part was sort of the end-to-end piece of the protocol that lets you put the things back together again. 'Cause these nets could be all different, and one net might take big packets and the next one might want little ones, so you have to chop these packets into little pieces and separately address them. They may take different routes and you'd have to figure out how to put them together and where the pieces go at the other end, and so that's what the protocol did. You might get duplicates, things could be retransmitted, they could come out of order, things could be missing, and so the protocol really was intended to deal with that. And there was a whole effort to really think that through as a kind of a logical framework for putting together networks independent of what the actual networks were, and I think the beauty of that original design was it seems to have accommodated virtually every kind of network that's come about since then and allowed it to be part of this federation of networks because the architecture was not about the underlying net, but about the means by which they could interact with each other. So that's why it was a separate activity.

Bob Cringely: It's amazing how many protocols and technologies have sort of come and gone, and TCP/IP is so resilient. You know, I remember it was like X.25 was going to take over the world, and didn't. You know?

Bob Kahn: Well, a lot of people used it and it's probably still around in a few places, although it was basically a virtual circuit connection arrangement. The thing also about the Internet was that it wasn't done as an abstract development, because when I got to DARPA one of the first things that I did was I got involved with the creation of two other nets. One network had actually been started a little earlier, but more as a kind of modem connection between nodes on the ARPANET, and we turned that development into kind of a stand alone net with separate interfaces and kind of gateways between them. They are called routers now. And another effort to build something called the packet radio net, which is a kind of forerunner of today's CDMA technology, if you're familiar with that.

The packet radio networks, they involved nodes that were maybe a cubic foot in size. They weighed about 50 or 75 pounds. You could lug these devices. You certainly wouldn't put them in a shirt pocket.

Bob Cringely: No.

Bob Kahn: And in the midline 1970's they cost about $50,000.00 each so it might be the equivalent of maybe $250,000 or $350,000.00 today. So these were very expensive units, but I believe, and history will probably show that if this was not the very first, one of the very first to use embedded microcomputers because the microprocessor had just been invented when we started that project. It was an Intel 8008 that was just become available. I guess the first one was a 4004. We ended up using something called the National M16, which was the first 16-bit micro that was on the market, and it also had spread spectrum technology embedded in this little box. Well, previously this was a technology that had been used in a number of applications. The ones I knew about where all military ones -- to deal with making signals work over very large bands so you could, for example, reduce interference or get the profile of the signal down.

Bob Cringely: Or make it resistant to eavesdropping.

Bob Kahn: Or jamming. Right. So we ended up demonstrating how this could actually work, and we had gotten surface acoustic wave devices from - I guess Texas Instruments built the first versions that we used and some more advanced versions from Lincoln Labs. We actually showed how this could all work and we fielded these units and we built small networks of something like 30 or 60 nodes, and the technology worked. So we had packet radio network ran at something like 100 to 400 kilobit range. We had the ARPANET, which was at that point a 50 kilobit-per-second leased line network, and then we had this packet satellite net that we had created on Intel Sat 4. And so we had three different nets, three different data rates, different packet sizes, different interfaces, and we had a very tangible issue of how to put them all together.

: And that was really the genesis of the Internet project.

Bob Cringely: Ah. So you had networks that you needed to integrate.

Bob Kahn: Yeah. It wasn't an abstract theoretical thing. We actually had a real test.

: I might point out that the way the Internet piece of it actually got started was more an integral part of the packet radio program because you couldn't carry around computers with you in those days. They're all million-dollar machines. You needed big air-conditioned rooms. They were as large as a very large 30-ton air-conditioning unit might be today. They were big units, and even memory was in - a megabyte of memory was a huge rack of cabinets, not a little chip back in those days. You couldn't carry them around with you, and if you wanted to get computing on a packet radio net, you really needed to connect to some of the net that had the computers on them, unless you were willing to compute off some trivial device.

So, there was good motivation to have the capability, but the way DARPA typically worked back then was you needed somebody that could make use of it from the military side. I mean, normally what DARPA did was to try and meet the real needs of the Defense Department going forward, so here we were in a situation where the military basically had no computers. Maybe they had a few for accounting purposes or for some specialized purpose. They were generally batch machines and they crank for hours and they spill out line printer output, but they didn't have any timesharing systems to speak of, or that really hadn't happened yet. This was, again, think about the early '70s.

So they didn't really have any need for networks to hook up to the computers they didn't have.

And therefore they didn't have the need for something to interconnect the networks that they didn't have. If they didn't have any, they didn't need to connect two or more of the ones they didn't have for the computers they didn't have. So it didn't really fit the traditional mold of what would happen, and so we were able to get it started sort of under the radar screen, just doing it as part of the packet radio program because that program needed it to make the radio network work, and then eventually it took on a life of its own from I would say roughly 1975 on.

Bob Cringely: Wow. In retrospect, in developing the TCP/IP protocol, would you have done it differently? Is there something you wish you had done differently?

Bob Kahn: You know, that's a question people often ask, and it's very hard to answer because you have all the retrospective of the history to deal with. I mean, how do you think radio and television would have been changed if we had semiconductor technology and lasers back in 1900? You can't run that experiment.

Bob Cringely: Oh, yeah. I know. But it's my job to ask unfair questions.

Bob Kahn: You know it would be different, but it's a totally unrealistic experiment to try to run because it's just the context is all different. I think we did a reasonably good job. It could have been better in some ways. Yeah, probably. I can give you some examples of things that we chose to do back then that in retrospect were pretty silly. We knew that we needed more address space. The ARPANET had 16-bit addresses. The way you would address something was you spelled out in the 16 bits which machine you wanted to go to. Well, guess what? They were all one net. Right?

So you need to say what net, but you had to say what node of the net to go to and then you had to say which wire out of the net. They were four wires so you can connect four machines basically. And so that's what it was. It was sort of like two bits to say which of the four wires and then another I guess six bits to say which of the 64 nodes that you could handle, and so those were using 8 of the 16 bits. And maybe there were a few more used for other minor purposes, but to first order that's what happened in the ARPANET. We say, "Well, if we're going to go to a larger set of networks, we need more bits," so we went to 32 bits. And our thought was -

Bob Cringely: That's enough.

Bob Kahn: - this is going to be good for the indefinite future. I mean, what we did was we took the 32 bits and said we'll take the first 8 and use that to designate which net. Well, if you figure out how many combinations of 8 bits you can have, there are 256 combinations, and we figured, well, AT&T would probably have a net of their own and maybe the defense department. That was two. Maybe there'd be one in Europe. That's three. We figured, well, maybe there'd be one in Japan or Asia Pacific or something. That'd be four. We double it. That's eight. Let's double it again. That's 16. Double it again. That's 32. And no matter how we could calculate it, 256 seemed to be more than we'd ever need. But what did we not think about? Well, we hadn't contemplated the personal computer being generated. We hadn't been thinking about local area nets that connect lots of computers in one building, for example, but within a very short time after that, all of that hit us. Xerox part came out with their altos and the Ethernet came right on the hills of it, so very quickly we realized that 8 bits wasn't going to do it. And so we had to kind of rethink how we were going to deal with that. We still kept the 32 bits. In fact, there are still 32 bits today. It's called version four of the IP protocol. IPV4. But people are now trying to move toward a 128-bit version, which is called IPV6, and it's hard to make those transitions. Anytime you move from one regime in which all the pieces sort of work together and you try and change some pieces then you've always got the issues of backward compatibility and how do you make the transition happen in a smooth fashion. So that was one of the things that I think we didn't get quite right initially, but we had enough flexibility in the system to cope with it.

Bob Cringely: Sure. Oh, yeah. The fact that we're still -

Bob Kahn: We're still using it.

Bob Cringely: - functioning with 32-bit addressing.

Bob Kahn: But we knew within six months to a year after that that 8 bits wasn't going to hack it.

Bob Cringely: Really?

Bob Kahn: But we thought 8 bits was perfectly fine when we did the original work.

Bob Cringely: (Laughter)

Bob Kahn: I'm reminded of the comments that people often make about the origins of the computer where it was maybe Watson of IBM had prognosticated that we'd only need a handful of these for the whole country, or in the early automobile industry when people thought that you'd only need a handful of automobiles to handle the needs. So, you know, sometimes these early views of what you think is reasonable can be overtaken pretty quickly.

So those were some - we thought about security right early on, but we knew that if we had to go figure out how to put security into the network, in the sense of real security from a military point of view, we never would have gotten this network built. It would have been so cumbersome and we wouldn't have been able to deploy them in the universities. It just would have been a nonstarter, but today people are still struggling with issues of security, and not necessarily in the military sense, but just in the sense of knowing who you're talking to and knowing what sites you're connected to and so forth. So I think some of those issues will get resolved over the course of the coming years, but that's one of the things we probably could have attended to.

Bob Cringely: Well, how could you have know? If you were sitting wherever you were sitting figuring this out, you didn't sit around saying, "Well, I wonder if someone's doing to do a denial of service attack on our network." Do you think about things like that?

Bob Kahn: Well, we weren't. We weren't focused on it. I have to say it occurred to us, but we were more concerned with - I think we were more concerned with how do we make this darn thing work.

Just to give you an example, when I was doing the original ARPANET one of things I was worried about was what I called deadlocks in the net. Too many packets showing up at one node so that the buffers got all filled up and maybe the buffers that the nodes nearby it got filled up or something so that nothing would move. You'd end up in a deadlock situation. Well, that's essentially a denial of service kind of thing, but happens naturally, and of course, a lot of the concerns that were eventually to come to pass were reviewed as not being realistic to have to worry about. Like worrying about in this room maybe all the oxygen molecules will just happen to wander up in the corner and we'll all suffocate.

Bob Cringely: (Laughter)

Bob Kahn: Well, you can worry about that and therefore require every room to have a fan and just get the air molecules circulated, and in some environments that's not actually a bad idea, but as a practical matter nature seems to do a pretty good job of keeping the oxygen molecules circulated in most environments. I don't know of anybody that's ever suffocated from the air molecules just happened to randomly go into a corner of a room.

Bob Cringely: No. I think it works just the opposite. Oh, never mind. So, we've already established that you had a hard time imaging the network being as big as it is. Was there a point where you came to adjust that feeling to think it's going to be everywhere?

Bob Kahn: Well, remember. We were thinking of this as a research experiment.

How do we connect computers, get them to work across multiple nets? It was a technologist's challenge. We weren't worried about how big is it going to get because we weren't trying to grow it necessarily on day one. Now in time more and more sites wanted to be added. How do we get to participate? And suddenly it started to develop a kind of a life of its own, and a growth rate of its own. And so somewhere along the way we clearly realized that this was getting to be a bigger thing and it would need to be managed in a more serious way. I used to create IP addresses and write them on index cards in my pocket and today we have whole institutions and the world fighting over who should be able to do that, and it was something - somebody'd call up and say, "Can I get on the net? I need an IP address," and we'd write one down and give to them and that was it. Nobody even thought that there was a business there, and today it's probably bringing in a few billion dollars a year.

Bob Cringely: So was there a box where the index cards went?

Bob Kahn: I never kept more than one 'cause when it got all filled up I asked - there was a fellow at USC named John Postel, and he had been keeping track of the host names for the ARPANET, so when a new site came on we gave it a name like MIT or UCLA or whatever, and John was sort of the - I don't know - it was almost like a green eye shade job in those days, but as it got bigger and bigger it became something to be staffed and managed. And John really became the kind of - well, he was the conscious of the net in the protocol sense. Is this a good thing to add? Is it not a good thing to add? With regard to things like domain names and IP addresses and all of the things surrounding it, he began to play a lead role. And he did that for some 20-odd years until it became almost overwhelming in the light of all of the issues swirling around - lawsuits flying about who could have what domain names and so forth, top global domains, and finally that led to the setting up of something called ICANN, Internet Corporation for Assigned Names and Numbers, which is still housed in the same building where John was at in Marina Del Rey. My next card never got to two. My numbers never go to two, so there was no box for it 'cause all I had was one of them. When that filled up, John picked it up and took it up. But we didn't have a lot of entries for the first few years. We had a very small set of research participants, and so I could probably have kept the list in my head.

Bob Cringely: Of course, a lot of the expansion of the net came from when it was opened up from the research community into - that was what? 1987?

Bob Kahn: Well, it occurred in two sets. When I left DARPA at the end of '85 NSF had already expressed serious interest in getting into the networking business. Eric Bloc had taken over as Director of NSF in 1984, I believe, and he was very interested in that, mainly to expand the power of networking to the research and educational community more broadly within the US and to some extent around the world. NSF ended up playing that role. DARPA decided to get out of the lead role of networking right in the middle of the 1980's. NSF really picked up the mantle. They created something called the NSF net, which was a higher band with net. They went from the ARPANET speeds up to 1-1/2 megabits, a factor of some 30 higher, and they used that to link together many different parts of the national, and it was a very ambitious project. A gentleman who led that over there was a fellow Princeton graduate. He had, in fact, the same thesis advisor at Princeton.

Bob Cringely: Which one?

Bob Kahn: John Thomas was his name.

Bob Cringely: I mean, which thesis? You had two. Bob Kahn: This was the first one.

Bob Cringely: Okay.

Bob Kahn: And this fellow at NSF was named Steve Wolfe, and Steve was really the architect of sort of moving it from this research phase into a much more operational phase, mainly for the university community and research community. He did a pretty remarkable job over there. He had help from a lot of people, of course, and managed to work the politics of the government. He got people from the Department of Energy and NASA and others eventually to pitch in, along with DARPA, of course, which had been there before, and by building on top of what we had already started with DARPA it really made the Internet as we know it today come about.

Now, it probably would have stayed as a network for the research community or maybe for research and government milieus had not there have been an effort that was spearheaded by a Congressman named Rick Boucher from Virginia to open up the NSF net to broader usage. I don't know exactly where the bill called it commercial use or not, but normally a government facility is used for government business, and this allowed NSF to open it up for other activities, so some of the commercials nets could now connect to it and traffic could flow over both of them without some kind of a policy on acceptable use. That's really what opened it up to the broader community, and it happened just about the same time as the world was discovering the World Wide Web. So the web browser was sort of a very visible entity by '94 or '95, but by '93 it was first beginning to become available in lots of places mainly due to the efforts at the University of Illinois, which incidentally we had been involved with to some extent because we had funded them to build an early version of one of the point click browsers.

But they used it mainly to connect to the supercomputer simulations, and some other people there then took that idea and used it to connect up to the world wide web, and that's really how the modern web browser as we now see it was created. There had been other efforts by other people and probably including Tim Berners-Lee tried to do similar things before, but this one really got some traction and took off. And of course, when Netscape was formed in, I guess, '94 sometime it just caused the whole thing to explode, and I think to Jim Clark's credit, who with the founder of that, Mark Andreesen, who left from Illinois, they really put that on the map and made the web a premier means of information access today. Not the last word, but it's just like the ARPANET was the first word in computer networking, the web is really the first word in network-based information access in a broad sense, and I think we'll see that evolve over time as people get more and better ideas.

Bob Cringely: Do you view this as your baby?

Bob Kahn: What's the this?

Bob Cringely: The Internet or the protocol. I guess what I'm wondering about is you have to and so it seems to me if you have some kind of paternal feelings about this, you must be concerned about how it's growing up.

Bob Kahn: Well, you know, people all the time write about the Internet and very often you'll see a headlines entitled something like Paternity Suit Over the Internet because you'll find three or four people arguing about who did what.

Bob Cringely: Oh yeah. The so-called fathers for the Internet.

Bob Kahn: The fathers of the Internet. I would put it this way. We've had a major success on our hands. Clearly. Just coming back from the World Summit in Tunis a week or two ago. Every national in the world has discovered the Internet. I don't think it's going away anytime soon, and maybe this form of intercommunication will be here to stay. Just like the interplay between cities took place and we moved away from localities to a broader global economy. Many people were part of that. It certainly never would have taken on the current strains of the technology if it hadn't been for all the investments made by companies. That wouldn't have been made if we didn't have the personal computer invented, if we didn't have the software to go along with that, if the semi-conductor industry hadn't enabled all of that. If we hadn't had the legacy of what came before us, I couldn't have done what I did.

People had been thinking about networking before. I was not the first to think about those ideas. Len Kleinrock and Paul Barand were early writers on the subject. Glen did a Ph.D. thesis on managing networks. Paul had looked at the issue of communication in the nuclear environment. How would you survive nuclear threats? His work started out by saying, "Well, maybe we oughta link the AM radio stations. Some of them will surely be able to survive, and maybe if they can send messages and they can get picked up by another AM radio station we can relay them around." This was an era before we had really VLSI technology. They were hypothesizing discardable electronics, but nobody quite knew how to do that in any reasonable sense. We didn't have small enough computers. So implementation-wise, it was pretty unrealistic, but they didn't have a Jules Verne-like issue of having to wait a very long time before you could take an idea and cause it to become real. So when DARPA started the ARPANET activity in the late 1960s it wasn't that far away from when those ideas were articulated.

But, you know, I think what DARPA did was kind of a fresh look at what they did, and of course, I had been starting to work on those ideas at BB&N independent of DARPA, and when I learned what they were doing it seemed a natural fit, so I became part of the process. So having been involved in the design and implementation of the very first computer network, I sort of feel a connection to the very early stages, and then having developed the packet radio net and cause these three nets to then come together sort right at the birthplace of this whole idea. Vint and I worked together on the protocols very intensively. I brought him into the mix because I knew a lot about the communication side of it, and he knew a lot about the computing piece of it and that was the fit we needed, but it would probably be a disservice to people like Larry and Landon and others to sort of say that their work wasn't relevant to this whole run-up because -- if you take a look at what was on the ARPANET it was a microcosm of what the Internet later became. People were communicating between different computers and all of the issues that we later had to deal with at the applications level existed on the ARPANET. It's just it was one net instead of lots and lots of nets. But what made it propagate and flourish was the ability to have lots and lots of nets work together, so it wasn't one company in Cambridge that was running the network for the world, it was everybody doing their own thing in this open architecture environment. So that's the real contribution I think I made to the Internet piece was sort of understanding this sort of open architecture environment, understanding the framework that could allow for these different nets to federate and specifying with Vint then the details how those interfaces would actually get implemented.

Bob Cringely: What are you doing today?

Bob Kahn: Having a wonderful interview with you. (Laughter)

Bob Cringely: Other than that, you still wear a tie. You're a medal winner we've just a learned, and so what responsibility do you have or you feel about this area that you were so involved in?

Bob Kahn: Well, let me mention a couple of things. First of all, I'm still involved to some extent because even though I haven't been working on the hardcore issues of the Internet now for quite a while we still have the responsibility for the standards process for the Internet, although were just about to turn that over, hopefully within the next month, to the community of researchers who have been essentially evolving the Internet. So one of the issues with the Internet was how do you take something that's open architecture where lots of different manufacturers can build things, where they all have to somehow work together, and allow it to evolve? It's one thing when one company builds everything. They can figure out a phase-in strategy to change over everything, and so we created something called the Internet Engineering Task Force a number of years ago. It really started out of some efforts at DARPA in the late '70s that Vint and I put together, and it sort of evolved, and Vint's still involved as Chair of ICANN to this day, but he may step down as that within the same time frame. I'm still -

Bob Cringely: Is the Internet Engineering Task Force going away then?

Bob Kahn: No. It's going to stay except that instead of the responsibility staying with us we're going to hand it over to the actual community of the thousands of folks out there who are responsible for it, who've got their own structure of how they're organized. What we did back in the early 1990s was we sort of separated it at the administrative part of it from the technical part and said, "Let's let the community be responsible for the technical part, and we'll retain the administrative part." Originally we had both responsibilities by virtue of a cooperative agreement from the National Science Foundation.

So that's happened gradually over time and so it seems like the right next step. The thing that I've been most interested in for the last 10 or 15 years was actually the work that we began here at CNRI back in the mid-1980s. I'm trying to understand sort of what the network might look like in the future if you were worrying about managing information instead of just moving bits. We came up with the notion of mobile programs in the net. We called them knowbot programs where knowbot was a coin term meaning knowledge robots. So these would be programs that move through the net looking for information and be able to do computation and bring things back. We've seen a little bit of that happen in the form of -

Bob Cringely: Spiders. Things like that.

Bob Kahn: - spiders and various crawlers, but this was a much more powerful notion, and as part of that work we broke it into two different pieces, one of which was sort of the piece that deals with mobility and another piece that deals with just the management, assuming you know where the information is. And I call that latter activity, although it really bridges both, digital object architecture. It's actually won some awards. It got the 2003 digital ID World Award for combing innovation with practical reality. It was an attempt on my part to rethink the Internet architecture around managing information. So if I wanted to sort of describe this in a nutshell, if you were to let all the information you cared about go onto the Internet, would you trust that as a going forward kind of activity? It's one thing if you trusted it to a place to keep it like in AOL or a Google or a Yahoo, or an IVM or something like that, and many people do, or maybe they trust it for themselves, but if you've got it on lots of machines and those machines can be in different places or the publishers are creating stuff and they've got it here and they move the content there - how do you keep track of all of the stuff when this material can move from place to place. And how do you manage it over generations of technology so that you don't run into the situation where you click on something and five years from now or 50 years from now you can't find it anymore 'cause links are broken.

So I came up with this architecture that is a digital object architecture which really has like three or four different components. One aspect of it is the digital object as the main lingua frank as opposed to just moving packets. Number two is a resolution mechanism that allows you to take identifiers to these objects 'cause every object has to have a unique identifier and be able to resolve it. Information about the object, call it metadata if you like, that among other things might let you find it, authenticate it, or whatever, or understand terms and conditions for using it. Third, more general -purpose, means of storing these objects called repositories, which is not a specification for a hardware or software storage system. You can use anything you like - any commercial database system, any flat file system, memory chips, whatever, where they can be in motion. And finally, registries of one sort or another that allow for some kind of authentication and metadata substance in the large -

Bob Cringely: Is this being implemented?

Bob Kahn: Yeah. We've implemented all the components of it, and we've had it on the net. The resolution system is called the Handle system. It's been on the net for over ten years. It is the premier resolution system. The publishing community is using it widely, and most of the electronic journals that are available on the net or that will be available in electronic form are now using them for among other things references at the back, 'cause they would like you to be able to take these journals 100 years from now be able to click on those references and still get them, even though they don't even know where they are, who owned them, and they may be in lots of different places.

I actually think that the power of the digital object architecture is potentially as important as the power of the Internet framework that we had originally developed. Because it's dealing with the substance of what's important to most people in their daily lives. I mean, it's one thing to be able to communicate with others. We've known about that for a long time, but the ability to manage your medical records or financial records, your personal papers, the things that you care about, they're not going to be publicly available for the search engines to go find. If you go and try and locate Hamlet, it's not clear that through a typical search today on the net you're going to necessarily find the text. You might find all the articles written about Hamlet, but you really want to go to the end thing. If you want to find your medical record, you want that record. You want to know it's your record, especially if a doctor is going to rely on it.

: So the whole object of that architecture is to deal with these kind of things so that you can literally manage your information over very long periods of time, and once you've made the investment in the information, what you'd really like to do is to know that that information doesn't have to be recreated every time you change your software, every time you change your hardware. In some sense it's flying in the face of a lot of the business culture, which would like you to do that, and for many things that's fine. I mean, I'm totally in favor of buying upgrades and whatever it is that keeps that industry vibrant, but once you've built your address base, you really don't want to have to redo it every time just because somebody's got a better data base system for retaining things. You want to be able to put the objects from one thing to another once you've created it and not have to keep recreating it over time. The same thing is true of data structures. If you're, for example, in the electrical property area you create a literary work or a musical work, you don't want to have to keep recreating the same literary work or the same musical work just because the technology has changed. You'd like to be able to have that expressed work in whatever the form that your original data structure was be preservable over time and manifested by different kinds of technology and platforms.

Bob Cringely: To be able to just refer to it.

Bob Kahn: That's right. Unlike object-oriented programming, which was a major construct in the computer science arena where people tried to hide the need for worrying about that detail from the programmer so they could write programs more easily and make the flexibility of that easier. For so many in the intellectual property business those data structures are their works. They represent their works, and that's what they want to manage, and so a method applied to an object in the traditional computer science sense is something that the owner of that would like to license -- rather than just sort of have it built in. And so you get a different view of how to manage information and data structures in this model.

Bob Cringely: But you do embrace the concept of inheritance, for example, like in object-oriented programming. I mean, that's the whole thing that allows this legacy data to flow forward, isn't it? Or am I just being stupid? (Laughter)

Bob Kahn: It's very important that when you create derivative works that you be able to understand either where they came from if the requirement of the original owner is that you maintain that. One way to do that is by essentially concatenating information in the identifier so you know what the original source was, but there are a variety of ways to do that. In many cases people don't want to preserve that. They say, "Okay. We'll write a contract with you." Maybe it's offline. "We'll give you the right to take our work and recreate it any way you like, any derivative that you want with our approval because that was what our contract called for, in which case it's your fresh start growing upon what I had done or somebody else had done," and both of those models and probably others need to exist in this world. But we took a look at this right from day one from the notion of managing intellectual property. We branched out from there. So you not only have intellectual property in things like literary works and musical works and audiovisual works, things like that that I mentioned, but you've got digital objects of all kinds. Some of those digital objects may not be intellectual property in the traditional sense - like they could be contracts, they could be mortgage bundles, they could be things that have value or just things that you have rights or interests in in some other form. And you want to be able to allow all of that to happen. If this were, for example, in the military, maybe it's information that's constrained by other means of access -- requirements or access. Maybe it's in a corporation. The requirements of access are controlled by company policy, and you want to be able to allow those kinds of access constraints to be built in right at the object level, or maybe pieces of the object, elements of the object themselves. So I think this is the kind of issue that companies, organizations, and individuals, all kinds are going to need to worry about in the future. That's why I'm so excited about it because I think the potential is there for this to cause just like the Internet was a federation of all different networks that were all different, this architecture allows you to federate information systems that are all different, including potentially those in the future that we don't even yet envision.

Bob Cringely: Wow. So in a sense nothing ever goes away then unless you decide for it to go away?

Bob Kahn: Well, it's your choice. You can have - I wouldn't adopt the situation 'cause I don't tend to throw away most of the things that I collect except that they're just trash, but you could have a model that says throw everything that says "Throw away everything after six months." You could have that model, built it in, and the system will keep only the last six months and everything else might be gone. Now, you may do it through some service provider that says, "Well, regardless of what they tell me I'm going to keep it anyway, because memory is so cheap I might as well," but then you better worry about what happens if you wanted to get rid of it because you didn't want anybody to ever find it and then some court asks you to produce it and they suddenly found that this guy, the service provider, had, in fact, actually kept it. You didn't want them to. So who's going to really decide what happens? And do you write that into detailed contracts? Do you remember all of that? Or you could say, "Keep everything forever, " or you could say, "Get rid of things according to the following style sheets," or you could say, "I'm going to write a program", or maybe some company writes a program that allows you to instruct it on what to keep and what to get rid of and now you're at the vagaries of this program, but the program might have thrown away something, or it might not have, and you may not know until you actually go and try and look for it and see if you can find it.

Bob Cringely: You know, I had a guy who came to me and he said that he wanted to keep all his records, everything about his life that he was doing (He was sort of a famous person.) and he said, but he didn't want anyone to have access to it because it might be embarrassing to him and what could he do? And so I suggested that he encrypt it all with like a huge 1024-bit key and then throw away the key. And then by the time they developed the technology to crack the key easily he'd be dead.

Bob Kahn: Well that's if you want it discovered, and if anybody's willing to take the time to actually do it.

Bob Cringely: Yeah. Exactly.

Bob Kahn: It might not be a bad idea if there's certain information you think you would like to know. For example -

Bob Cringely: Or society deserves to know, but not in time to put him in jail. (Laughter)

Bob Kahn: Okay. I think I'm going to pass on that one other than to say that one of the things people have thought about - there's a greatly expanding amount of information that we collect - is to sort keep a light history of the person. Maybe wear a little camera on your lapel or in your hair, something that sort of records what you do every moment of every day of your life.

Bob Cringely: Ted Nelson does this.

Bob Kahn: Today? He does?

Bob Cringely: He's been doing it for 40 years.

Bob Kahn: Well, there may be times he wanted to turn it off, but that's a different -

Bob Cringely: No. Ted doesn't turn it off.

Bob Kahn: Never turns it off?

Bob Cringely: No. Pretty much not.

Bob Kahn: But if you could do that and you have a complete history of what you've gone through then somebody could experience your life. Now maybe this is a person who becomes President of the United States at one point in time and the country would like to know, but that person doesn't want you to know during his lifetime. Maybe that would be an interesting way to find out. On the other hand, most of the really interesting stuff about people are, let's say videos that show them as part of the scene, so you'd almost need to have the need for somebody to be attached to you as an alter ego essentially recording what you do external to you. Not from you looking out, but from them looking at you.

Bob Cringely: That's what Ted would like, (laughter) but he can't afford it.

Bob Kahn: Well, who knows? Maybe if we come back in a few hundred years, maybe it's less than that, maybe a little fact totem of sorts that comes with you. Maybe it rolls on the street like a Segway or a maybe it sits on your -

Bob Cringely: It's your knowbot?

Bob Kahn: Maybe it's got butterfly wings and it flies off and photographs you from ten feet away and sits back on your shoulder when it's no longer interesting. Who knows?

Bob Cringely: Who knows? The people with real power are the ones who are allowed to turn it off.

Bob Kahn: I suppose. Yeah.

Bob Cringely: Bob, thank you very much.

Bob Kahn: You are so welcome.

Be more connected. Help support your PBS station and projects like NerdTV...pledge online.

© 2005 NerdTV LLC and PBS. Offered under the Creative Commons license.

Privacy Policy | Terms of Use | Published on September 6, 2005