Interviewee: Lawrence G. Roberts
Interviewer: Bu Zhong, Lee Yin
Date: August 03, 2017
Transcriber: Fan Yuanyuan
Lawrence G. Roberts is an American scientist who received the Draper Prize in 2001 and the Principe de Asturias Award in 2002 "for the development of the Internet".
As a program manager and office director at the Advanced Research Projects Agency, Roberts and his team created the ARPANET using packet switching techniques invented by British computer scientist Donald Davies. The ARPANET was a predecessor to the modern Internet. In this interview, Roberts talks about his childhood, his family, his study experiences, his career development, his leadership at the ARPANET as well as at subsequent companies that he helped co-found, and about networks security and leading edges on the internet.
BZ: All right, good.
LR: Okay. Well, my name is Lawrence Gilman Roberts. I was born in Westport, Connecticut in 1937, and my parents were Dr. Elliot J Roberts and Dr. Elizabeth Gilman Roberts. She was… They were both PhDs in chemistry, and the whole family got PhDs. Um, the, the… I went to, uh, Staples High School in Westport, Connecticut. Um, and I then went to MIT, after that, uh, for my undergraduate and my graduate work. My master’s was mixed with my bachelor’s. I went for one, and, and, then I finished my PhD, uh, while I was working at Lincoln Labs as well. Yeah.
BZ: How many total years, you know, you studied at MIT?
LR: Well, I was, uh, from '55 to '63; so it was about eight years (BZ: Okay) total, including the PhD.
BZ: Yeah, I didn’t realize that you were born into a parents who all, well, you know, already have, you know, both of them have PhD degrees in chemistry. Um, I can understand maybe one parent got a PhD; now, your mom does a PhD. So, how, that's, you know, make you different from other kids at your age? Most of us, we don't have a PhD parent at all.
LR: Right. And so it's certainly enforced the concept of being in science, being, doing something in science, and choosing the right science field and doing something in it and getting my PhD, um, which I did. So the, the, the, um, the choice at that point though was which direction did I go, chemistry or physics or what? And I saw electronics as being the new thing, and so I wanted to get to the new thing. Um, and so, I pursued that at MIT, and it worked out very well. Um, when I was in my senior year at MIT, the the, the TX-0 was brought to MIT. Uh, it was the first transistor computer in the world. And it was, uh, the prototype basically of the DEC [Digital Equipment Corporation] PDP-1 And, it, it was given to MIT to work, but nobody knew about it, except I found out about it, I spent 700 hours programming in that, in the first year online, uh, uh, myself. And that was totally unheard of. Because most people were using the big batch computer, the IBM main frame. And they get a few minutes; uh, uh, you know; I know the computer at best. I did a little work on it and I had installed a drive. So, so, it was a pain to work with the IBM machine for me with punch cards. But anyway, uh, the TX-0worked tremendously well. I did the first neural net program, doing character recognition of handwriting, uh, probably not well-recognized in my literature as one of the first, and perhaps the first, uh, neural net program of multi-layer, a program that we were all doing now on artificial intelligence. At that point in time, the artificial intelligence group at MIT thought that was not part of the mainstream. It was (chuckles), it wasn't what they should, I should be doing. But I figured that was a useful thing to try and do which I did in my senior year. That was just the paper, one paper, I published.
BZ: Very good. You know, uh, before we go to your, um, uh, MIT years-- that`s absolutely amazing, you know, 700 hours and working on PDP, and, you know, that kind of thing there, instead of using IBM punch cards and the computer. So, let's go back a little bit there. So, uh, when you were young, what kind of influence you get from your parents? You, you have siblings, and you have…?
LR: I have two sisters who are older, and they uh, uh, they didn't influence me tremendously, but, you know, that, that was my family, and, I, my parents were tremendous influence in terms of uh, huh, doing things, working hard, and, and making things happen. My mother, um, since, the, since the world at that time didn't accept both parents working very easily, she went into, um, volunteer work, and ran the Girl Scouts and ran the United Fund. So, she became the chairman of, of these various groups, and ran them. And, so, I saw a leader there, as well as my father leading; he was the vice president of research at, uh, Dorr-Oliver. It was manufacturing to sell his equipment. And, so, he, he was doing, um, running a lab, which was near our house; so, I could go there and see him and his people.
BZ: How he feel, like, you choose, uh, you know, doing something electronic engineering instead of chemistry, um, he, he lets you choose what you want to do?
LR: Oh, sure. He let me to do whatever; in fact, I, I built the first TV in the house, uh, the uh, TVs were just coming out, but they didn't have one, and I built one, um.
BZ: What year was that? How old are you, you were, at that time?
LR: Oh…12 or so, I guess. I’m not sure (BZ: Oh, okay), um, it was, it was early and, and, I, I built all sorts of things, Tesla coils and induction heating; uh, we are now starting to use induction heating for, for cooking but I did that back then (BZ: Wow, my…) and all sorts of other things that I tried, but my, um, but because my father was a catalyst and had a bunch of chemicals around, I was, I was also able to, uh, build nitroglycerin when I was, uh, in first grade, in elementary school. And I, actually, took nitroglycerin to the school that [found / pound / held / bound…??] (unintelligible) on a rock. Uh, but I was, I was, I failed in terms of the temperature that I cooked that at; it, it wasn't quite right, um; I did it better, later.
BZ: Go back, like, that you make a computer, I mean make a TV set, like a 12- or 13-year-old boy, that's quite a big news in your neighborhood, not only in your family. Is that right?
LR: Well, it wasn't much of a neighborhood. I mean, my, my, uh, house was on its own ten acres and there was nothing around. I had a friend that I could walk to not just about one, and, um, we ran around the woods and, and explore things. But it was, it was, uh, not really a neighborhood of a lot of kids.
BZ: Oh, okay, okay. I understand that's in Connecticut (LR: Yeah), that's in Connecticut there. So, how's the family reaction to this? Oh, suddenly we have a TV set in the house, were built by the youngest son in the family.
LR: Well, I mean, it was, uh, they just accepted it, it is what was happening. And, and, they didn't watch it much. Uh, it wasn't the big thing for them. We put it in the playroom and the kids watched it when they wanted to.
BZ: How big the screen is?
LR: Oh, it was, it was about this size, I mean, it was reasonable size too (BZ: Wow) that is. black and white back then.
BZ: Yeah, yeah, yeah; but still quite, quite a piece of work, you know.
LR: Well, I built a lot of things, and you know, radios and transmitters and other things that, uh, I had the material to do or found a way to do, and I bought most of, borrowed, and put together a lot of test equipment, so that I could, you know, had a sole scope, and everything else. So I knew most about electronics before I even went to school.
BZ: To go there. So it's a natural choice for you to go to MIT for your college?
LR: I applied to Yale, Harvard and MIT. And, and, my father and mother went to Yale, uh, for graduate school. And, so, I got accepted everywhere, but MIT seemed like the best choice.
BZ: Huh, and, you get supported by your family?
LR: Oh, yeah, my parents were totally supportive of it, and, and, and, in those days they, they could fund it, um. I don't know that it's so easy today for people to fund their kids, but they could fund it through the undergraduate years. And after that, I had a research assistantship which made over my expenses.
BZ: Okay, so, you your sisters go to college, went to college, too?
LR: Yes, my, my, uh, younger sister went to, uh, Tufts and became a professor at, at University of Boston, um, University of Massachusetts; she became a professor of biology. And my older sister, uh, went to, uh, Cornell and then, uh, became a teacher and so she (BZ: As a professor?), where she was a teacher in special ed [special education] and, and, so in, in, not in college, and, and, and she moved to with her husband around. So, she wound up in Colorado (BZ: Okay) teaching there.
BZ: So, as soon as you got into MIT, you began to develop, like, a strong interest in computing?
LR: No, no, because, MIT, we didn't, computing didn't exist, really. I mean, uh, the, the first I saw of computing was the, uh, 704 that they put in the basement and asked me to have the tape drive to it and the TX-0. There was, of course, other activity at MIT but I didn't see much of that. And then, I went to Lincoln Lab where the TX-0 was built and, and we started, uh, with the TX-2, the TX-2 was the predecessor of the PDP-10. And it was 6microsecond cycle, I mean, slow, by today's standards. But, even so, it, it was powerful enough to do virtual reality when, when I built that up with Ivan [Sutherland]. And, uh, and it's, it's amazing what we could do with programming in machine language at that point. History was to get very fast action. I could get the display in 3D working in rotation, and in any rotation I was on. So, it was, it was pretty effective. When I started at MI…,at Lincoln and they, they showed me to repair computer and said, well, there's a punch board here, plate panel, and that's the only way we have to do anything at the moment. You’ll have to build everything else; the, the operating system and compilers and assemblers, anything else. So, I built every, all the, all of the libraries for the system. I built this timesharing system and, the, um, and the compilers and assemblers and, and, so on, all from punch board and, well, then I read in paper tape and went from there. And, and eventually I had to add other equipment to it, like when we got to the point where, uh, MIT`s Marvin Minsky who wanted-- he was in artificial intelligence-- and wanted that some of the pictures that I had been reading into the computer at Lincoln, because I built up a facsimile, old facsimile drive to read in photos. So I could do my work, which was recognition of photos in 3D, and for my PhD thesis. And he, he wanted a lot of photos that I had, a lot of them were art photos that, had taken photographs of, and a lot of them, for, um, photos that he could work on processing. So, what I found was there was no way to move the data to him, because the TX-2 was independent; I mean, different than his machine. Even though he had a PDP-10, it was a different machine in terms of the architecture, well, compatibility. So, I had to, uh, build an interface for an IBM tape drive and, build, and put in an IBM tape drive and write a tape for him. So, he could read. Even though we, the only compatible media, which is one of the things that encouraged me to worry about networking, because we couldn't move data between the various machines. We couldn't do things; and all the different machines were all different.
BZ: Right, in those days, those kind of TX machines, how big were these?
LR: Well, TX-2 was the basement of Lincoln labs; it was bigger than his house, uh, uh, it was huge, it went on forever and in any direction, and it was bays of equipment, I mean of transistors and the first core, first core memory.
BZ: Right, so that's actually give you a strong desire to want to interact with other machine, with other people, that's something, like, and, come up to ARPANET?
LR: Yes, uh, see, I was, I finished my PhD in '63, and '64 I met with, uh, Licklider at a conference in Virginia where we were all just sitting around talking between sections, and Lick was saying that, what, what's the future, what do we need to do next, where, where we need to go. And he was pushing the concept of some sort of network to tie the machines together. And that, that was something that I, uh, found very exciting as, as a project. Because, although I could keep on working in graphics where I was ahead of everybody in the world, and had all the seminal papers, in fact, they, everybody quotes them all the time, you know, because these were the first papers I’m doing, 3D, uh, well, picture recognition, fixed picture tracing of edges and finding, finding the images, and then creating and, and matching it to a 3D objects in space, in, in the memory, and building up an object that I could display in 3D for many directions, dimension, any direction. And, so, I built the hidden line display of 3D display, which was probably the first, well, it was the first 3D in line display in, in history. And developed the math for that, which turns out nobody since the old German text, uh, had taught much about projective geometry and its math, and so they said, you got to divide by this and do that. And, and, it was really cumbersome mathematically; it was really a pain to have to do the divides all the time, where computers don't like divides. So what, what I did was, I turned it into a four dimensional matrix transformation, so that when I wanted to rotate the point of view from a different perspective point, we did in four dimensions. Everybody uses that now, that's the standard homogeneous math for 3D. But that was, that was probably one of the most referenced papers in that respect, because I figured out how to do the projective display, efficiently and easily.
BZ: So, we were, we were at, today, you know, there's a little bit argument about the beginning of the internet. So, when that internet really came into being? Some people will say that's back in 1969, there is a student, you know, began to use internet, to send a message from LA, UCLA; he’s at the lab. Um, what’s right in ityou know, when the internet really came into being?
LR: Well, the, the, the concept that I, uh, had… let me back up a little bit, because I got to fill in the history. Um, Licklider left, uh, ARPA and left Ivan there, one of my friends from MIT, whom I had worked with, and he gave me a contract to work on my prototype concept of how to do, connect, two machines as a network. And I programmed that, uh, with a team connected to a computer in, in the West Coast and learned that basically the telephone network which I used as dial up network was hopeless for data communications. One, the computers were bursting 15 to, uh, 15th of the time and the rest of the time they were idle, so that we're wasting 15, uh, 14, 15th of the time, uh, on the dial up line; and, secondly, the speed was low and the connection time was very long and everything was slow. So, I…my roommate also at MIT was Leonard Kleinrock and he had written his book on how to build networks, I mean what the concept of a network would be with topology and, and queuing theory, in particular. He was an expert of queuing theory. And he had showed that you weren`t going to lose all of the packets, if you buffered them, that, you could set a limit on what you could do, which gave me the confidence that I could build a network as theoretically. His was the research. So, I, um, I, was, then put together a plan in my mind of what to do. But, um, I didn't want to go and do something at ARPA which is what Ivan asked me to do, and then his, his follower, uh, deputy, uh, Bob Taylor also trying to get me to come, because that was what the group had decided that I should be the next director of the ARPA, uh, information technology office. And, the, the problem was I didn't want to get into management yet, uh, and, I was having fun, doing, working on the computer. But I knew that, um… so, he, he, he gets [Charles] Herzfeld to, uh, call the head of Lincoln Lab to say-- get him to come here, which he, he talked me into, eventually. And I went to ARPA in '67, met with, uh, laid out the plan, designed the network, and, um, generally, it, and, it took several years to design the network actually, because the topology is, nobody had worked on that much and there were all sorts of things to figure out that we hadn’t, uh, learned yet. And, so I, I designed the network and got it built, uh, in my mind and, and when I arrived RFP [request for proposal] for somebody to do the, the computer part, uh, chose BBN [Bolt, Beranek and Newman] in the end, for that, and chose UCLA where Leonard Kleinrock was now for [Network] Measurement Center, and got all of the universities that I was funding now for research, for computer science research, to be part of it, to, to work on it. And they weren't too excited to do that, they didn't want to lose their computer time to somebody else. But they went along with it, because they, I was funding them. And, so, uh, basically, there was very little interest at the time in the computing community to, to do this, but, um, we'll get back to that change very quickly as the network started. But in any case, I then got the, uh, network together in '69, which, uh, you know the, the date when UCLA finally got connected to SRI, and we had the first link up. And the links grew after that very quickly. I put in more lines to the East Coast and around the world, I mean, around the US, to build a network up. And we installed more of the equipment throughout the system, including ones that you could dial into, so, you could get to a computer somewhere else, even if you didn't have a computer, otherwise it was computer to computer (BZ: Right). But, um, so, it, it took between 1969 and 1971 for that to become a working reality, a working system that everybody could use efficiently and where there was real traffic. And at 1971, uh, I, um, they began, finally, finished the programs that were supposed to be done, that would transfer a file between people, uh, and, and, and, and as, as you know, the”@” sign was used to, to, to designate the person or designation of the address. And I found that totally, uh, ineffective because it was like a teletype. I, I, all the messages got streamed out on my printer. that was not all I could do; there was no way to interact, really. So, I wrote the first email program and in a text editor called TECO. It was, it was, uh, was the first email program that, where you could save a message, read them in different orders, just get the headers and, and, and save them and resend them and keep track-- everything we do today with email…
BZ: What's the email system called again?
LR: Uh, that was built on, it was called--- almost have to look up the name, uh…
BZ: That's not Pine, right? (LR: No, no, no). Pine is later on; Washington University.
LR: This is, this is, uh, the first one that existed; everybody started copying in different systems after that.
BZ: But, that year is 1971?
BZ: Okay, yeah.
LR: And so, with that, the director of ARPA, Steve Lukasik at that time, said, my, my god, that would let me communicate with all my people all over the world all the time, in real time, rather than, I mean, and get it the same day. So, amen. So that the issue was, um, he asked everybody in ARPA to use it. Well, the only other groups which were doing strategic, uh, defense and other things. And so that was one application. And also all of the universities, that TECO editor was available to everybody, so they could copy the program, really easy. So everybody had a copy of the program. And it was straightforward to, um, everybody to use it for email. Email became the biggest, uh, business on the network in 1971 and 1972. It was the major activity, um. But also there was a lot of other activities going on and, and people were building other functions over time as they learned to use the network. And even [Marvin] Minsky and [John] McCarthy who had stood/slept [u1] who had separated-- McCarthy had moved to Stanford-- could now write a joint paper and, and, they both didn't want this network but they found it very useful because they can now work on a paper together, and send copies back and forth.
BZ: Let me go back a little bit to these networking here. There's a key technology breakthrough; I think it is queuing, it`s queuing networking or queuing wiring?
LR: The, the fact that you have to queue all of the packets in memory and not lose them before you use them (BZ: Right), uh, was an issue of how many you could store, what the frequency would be, so the queuing theory said how big that buffer would build ( BZ: Yeah), and that that was what Len [Leonard Kleinrock] had worked out the theory for. And so…
BZ: Did he do that? We're going to visit him, um, tomorrow, yeah, you know.
LR: He’s the world's expert on queuing (BZ: Okay); he worked this out back, and, and wrote a book actually in, in 1964, to, about networking.
BZ: But it seems to me, and, many of you also contributed to the idea of the queuing make it, you know…
LR: Well, I mean, his queuing theory was, uh, was the math of how big the buffer was going to be (BZ: Right, right). We all knew about queuing and we all knew about-- anybody in computing in those days, knew about, uh, packets. In other words, you wouldn’t send a file as a file; you would send it in pieces, and, so, you can come check it. Uh, and the first thing I did in my first experiment between two computers was to decide on the length of the packet that could work through the telephone network. Works had huge data noise from relays going by nearby, uh, so, there was, there was burst errors. Burst errors destroy a packet, not, they don't, they're not a bit error rate there, they are a packet error rate and affect, when the burst occurs, the packets-- it`s killed. So you don't want too big a packet. So that I, I worked out in that experiment-- what size packets we should use; at that time, about a thousand…
BZ: So, you did that in Lincoln Lab?
LR: Yeah, I did that while I was at Lincoln
BZ: Ok. All by yourself or with some of your roommates or your professors, or …
LR: Well, no. I did this by myself. I mean my professors were overseeing my thesis (BZ: Okay), um, they didn't pay much attention, but I mean, they, they were overseeing it. I mean [Claude] Shannoon was one, [Peter] Elias was one, and Federle was one. [u2] They were the three advisers. But they didn't get involved in details. And I worked, I did talk to Kleinrock and, and I did talk to Ivan--those were the two-- Ivan Sutherland; those were the two people I tended to talk to about, about this type of thing.
BZ: But that's, and, you know, you, designed this, you think this through, that took a lot of time of you.
LR: There was a lot more time involved. When I went to ARPA in 1967, it was from then until probably in about 1972, by the time I had finished optimizing the, the, the topology, and the, the whole concept in working out the papers, writing papers about what the, what the effect was, and how it worked, um. But, in particular, I spent a lot of time working on the topology to optimize the cost, Because, uh, the, you had to look at the cut sets of all the lines and find out the throughput across in a cross section, and see how much data you could get from A to B, and then you have to work out the matrix of how much each user is going to be sending to where, which place and, then, you can figure out what the networks should be.
BZ: Right. I assume you've spent a lot of time, so much time in the lab, behind those doors, and might hide yourself behind those, kind of, big computers there. Anybody saw you on campus?
BZ (chukling): You always worked so hard.
LR: Not, not after my senior year. They didn't see me on campus.
BZ (chukling): You, you attended parties, or like, you know…
LR: Well, I mean, there were, there was events, but I didn't pay much attention, um, uh, but I worked with all the people, people there. When I was at Lincoln Lab, I had a team working for me, to, because we, I took over, sort of, arbitrarily. when I went there, Wesley Clark left and Wes, uh, was the leader of the group and he, uh, he and Bill Fabian left too, because they weren't allowed to work on cats, uh, that Lincoln decided that they couldn't do, uh, brain research (BZ: Right), at the same time to figure out how the brain works, so they could build it on the computer, uh, so they moved. And that was all of a sudden, and I was left with the whole group (BZ: Okay) and TX-2 sort of to myself. So, I ran just because no one else did.
BZ: Where did you go to lunch while you are working there? You bring your own?
LR: Uh, Lincoln had some sort of cafeteria, I think; I don't remember.
BZ: Okay, you sometimes could bring your own lunch box?
LR: Oh, I bring my lunch mainly, normally, everywhere. But, I mean, even today. But, but, but, but I don't know what I did exactly there for lunch. I didn't have a lot of, of, of socializing with people. I was mainly involved with, well, and I was, I was married by then and I was mainly involved in the operation of the group and my work.
BZ: So your brain is totally occupied for your work?
LR: No, and, and, yeah, I have to bring up the whole operating system for the computer as well as the, the, my thesis work, I didn't, my master's thesis was, ah, a picture uh, uh, coding, so, that I could reduce the bit rate and they used that patent for, uh, the moon shot. Uh, because all of the, all the pictures they were getting back from the moon, I had to compress. And this was the only compression that was logically possible in the logic of those days’ transistors. Uh, there's no integrated circuits (BZ: Right), so, it, it got built that way. The patent actually ran out long before anybody else would use it in much better algorithms occurred later on. This was, this was when using pseudo random noise to fool your eye into believing you were seeing a clean picture without, uh, without, um, bars, lines where the, where one gray scale late, late into the next. And, I got rid of those by adding noise and subtracting noise later.
BZ: Today, when we look back and like, you know, in the earlier 1970s or late 1960s there, we were appalled how much you really achieved, but actually behind those kind of achievements, lot of work, lot of, like, and, you know, hours and you may have seven hundred hours, and working on the machines (LR: I worked at home). Are you a lonely, um, scientist working there?
LR: I’m not, I’ve never been lonely (BZ: Okay, so you're…). And I lived and, and I worked at, at my house, um, and in the evenings as well as during the day at the Pentagon; I got everything uh built. I was also managing all of the R&D for computer science (BZ: Okay) so my program started out at 15 million dollars a year that I was responsible for all the computers science and research. And, overtime, by careful managing of how the budget process worked in our apartment, I grew it up to 50 million, um, in the years. I stayed about 6 years because that's all I figured I should stay anywhere.
BZ: So, in those MIT years, I know, and, you devote a lot of time to, to work. Do you have some hobbies and go to band, or sports or baseball, you know?
LR: No, I don't do much in sports, um. I have in the past and the various sports. I played hockey in MIT for a year, but nothing serious, um, and I don't, I don't go to the sports events, particularly.
BZ: But you find your love.
LR: Well, I, I’II certainly love woman, and my wives, and over time, I’ve had several wives. Um, and I love my work and love working with people on.
BZ: Okay, good. So, um, this, let’s come back there. So, do you agree, like, and, you know, internet was born in 1969?
LR: Well, clearly from that point on, uh, and, and there was only two computers tied at that point in, in early 1960, in late 1969, um, by, by1970 and 1971, certainly in 1971, the internet was totally functional the way it is today. I mean, we could do anything we are doing today, in fact the bandwidth was very good. Uh, um, we had 50 Kb lines, and the lines, er, provided the bandwidth we needed for our activities at that point in time. Moore's Law has grown everything. So, the network had to expand with Moore's Law, effectively, because, as people build bigger and bigger files, if you're, if you're, uh, I mean, today, when I’m working with my coworkers in my current business, um, uh, they send files that are several gigabytes, um, these are big files from the test data and then I analyze the test data. And, so, we're exchanging files that are huge, uh, and Google Drive allows you to do that. But the point is that the network has to be fast enough to move these big files. And it has grown over time, as I noticed over time, as I worked on computers and the network. I mean, I’ve worked on the networks since 1970, 1969, actually. Um, then, then basically the, the files keep getting bigger. And what I’m willing to do is things that I can handle on my computer. And, so, as my computer gets faster, I’m going to be building bigger files, and doing bigger things. And the network has kept up with that up to a point, and now it has sort of stagnated. I’ll get back to that later on, but there is a problem that in the future, because, although the trucks are running a hundred gigabytes, you can't send the file over a hundred megabytes across the network today, because of the TCP, and TCP is a great protocol. Um, but it, it, but it doesn't use any intelligence in the network. The network doesn't, the, the network drops back as if it doesn't like it, and it doesn't do anything to tell you effectively how much except to drop the package, and that isn't the way I designed it. When I designed the network with NCP [Network Control Protocol], the first protocol, uh-- Never Control Protocol (laughs)-- it, it was, it was, uh, my design was to have storage in every node, and check the subject before you forwarded the packet, and send acknowledgment back to free up the space for another packet, so that I was continually informing the senders how much space they, how much they could send, so they couldn't overload the buffers, and then I was sent checking every node. B time it got delivered to you, it was correct; you didn't have to use TCP (BZ: Right). And, so, TCP didn't exist at that time because it wasn't needed. I then…
BZ: So, when did you design NCP? That`s really reason, nothing control…
LR: NCP is, actually, was designed by my group of graduate students, uh, but it was under the expect that that's the way I wanted done (BZ: Right). I, I didn't do the details.
BZ: What year was that?
LR: That was in 1970; it got designed because it had…wait, 1968 and 1969; it got designed and was operating in1969 (BZ: Okay). It had to be there.
BZ: Okay, so, the name is Nothing Control Protocol?
LR: Network Control Protocol (BZ Laughs). That was, actually, Vint Cerf was one of the people on that committee that was run by Steve Crocker. And, so, Vint was familiar with it (BZ: Yeah, well, we'll talk to him in DC; yeah), but Vint, Vint decided later on after I left ARPA to, uh, convert everything to uh, uh, not store in the network, not have intelligence in the network, and not decide what, how much load you could have. But he decided fibers are fast enough as things were going forward in history. So, let's do it all from the edge with TCP, and, and have it try to figure the path, I mean, figure out how to, um, how, how fast to send. And that worked great for quite a while, um. It has a problem today because that's what's limiting us do a hundred megabytes.
BZ: I really wanna, you know, got your opinion about TCP/IP, you know, today, later on, yes.
LR: But, but the other thing I, I then did was after, after ARPA, I went and built Telenet which was the first public payload carrier for a packet switching. I mean, this is all packet switching. The, the internet, as a thing, is those computers that are interconnected as a common media, but there's private networks and there's and then we built, uh, in, when I, when I finished with ARPA in 1973, I formed Telenet. Telenet became a carrier, common carrier, actually. And, and offer services throughout the world, through other carriers where we provided equipment to them, actually, because we built the equipment to do the packet-switching. And this was X.25. X.25 was a protocol I designed from scratch which, uh, because I knew we couldn't have a network without a standardized protocol that you could say used XYZ to connect to our network and XYZ was designed and built for your computer, it was available, hopefully. So, um, I designed X-25 and went to the ITU which was then CCITT and, and, uh, worked with England, and France, and, well, England and Germ… and Japan and Canada and ourselves in France with the only ones interested, and we, we, we got the thing passed; there is a standard, very quickly, in, by 1975. And, and then, uh, that was what we started using them as the international standard for packet switching for the next 20 years.
BZ: Wow, that's interesting. I definitely want to know a little bit more; how about you have a coffee break?
BZ: So, I think. so many genius ideas, um, plug into your system and make it work so well. So, um, I believe you have so many highlights at the MIT. Nowadays, we look back, we always think of your ARPA experience, especially, ARPA network, and, became so well-known. Um, I, I like you tell us and, you know, more highlights, uh, you know, there. Let me see, how it is (checks the microphone). Yeah, we're recording, yeah.
LR: So, what was the question?
BZ: My question is your MIT highlights. You know, uh, you, you, how many years you worked for Lincoln Lab?
LR: Well, um, 1960 to 1966; so, about six years.
BZ: Six years from start, as an undergraduate student?
LR: I was, I was, I started as a… yes, undergraduate, basically, my, in my master's activity. I, I did my master’s thesis at Lincoln (BZL Okay) and then my PhD thesis, and also ran the group, um (BZ: Okay). So, let's go back to MIT. The first year at MIT was, um, uh, I got all As, and, and the people who did, they collected the 30 of us and put us in a special class, called it, instead of course 6, course 6B, uh, 6A was already used for industrial relations, but 6B was a new course; it doesn't exist anymore, it was, it was tried. I don't know that the professors had the time for it, really. But it had the main professors for the course teaching us as opposed to the other lecturers and so on so that we had the primary professors, and we had this small class of 30, and we had the, we, we were all very bright. So it was, it was a very easy, uh, process through, through there. And I, I worked with, uh, that was going to, direct to master`s now. And, so, that's what I did with at that point. and then as, of course, as, as part of that which is now starting much earlier at MIT, but we started getting introduced to people who would let us do experiments in the lab as opposed to just courses. And, so, I did all sorts of lab experiments, including one where we built in a sealed room with all sorts of, uh, door locks and everything, uh, electronically controlled from the next room. And I built a power supplies, and I built the, the, um, capacitor bank, and a huge, uh, I used to, to run a single jolt of power through a, a single turn coil to make the biggest, garish, uh, biggest magnetic pulse that was known. So, the professor really wanted to do the largest, uh, jewels is it drills?] that anybody had ever had, so we, we charged that voltage capacitors for days and then, uh, it was a room full of capacitors, I mean it was like this big room sort of just capacitors, and it then destroys momentarily through the, through the, uh, coil, through the single turn. And, it made a huge, huge, um, magnetic field, which, he was, that was his interest (laughs). I didn't have find that particularly exciting, but I did all the work. And then, I did another one where I spun the ball on a, a, in a high frequency rotating field in mid-air, so the ball floated in mid-air. And other things that I did, it's when I got into computing, uh, and got under the TX-0, so that (BZ: That's almost the second year and) that was the fourth year (BZ: The fourth year, ok). In the second and third year, I did these other things (BZ: Okay, good), so, MIT was, was a bridge, um, it worked; I mean, for me, anyway.
BZ: So you start your master’s program, and, you know, in the, in the fourth year, after four years...
LR: No, actually, it was a five-year program to include a master`s (BZ: Okay) so what was, we wouldn't, we wouldn't stop for, uh, bachelor’s, we got that along with the master`s (BZ: Okay, and, this, three more years for PhD?). And then, then I finished that in 1960 and 1960 and then 1963. I finished the PhD in three years. Some people take 10 now, but…
BZ: Okay. So, what are your years, after your MIT graduation? So, what, what did you do after that, after your graduation there, so where you started?
LR: As soon as I graduated (BZ: In 1963), I was at Lincoln (BZ: You go back to work there). So, I was continuous. I was, I was on the staff already. I was, I was being paid as staff member, by the time, I was…
BZ: We call that, and, a research scientist?
LR: Well, at Lincoln they just call a staff (BZ: Oh, okay). I mean they, they just don't have a lot of designations me. And, and, in fact, I asked the head of Lincoln at one point—“here I’m running the whole group, should I be a manager or something?” No, no, you're fine, just do it (both BZ and LR laugh). Staff is fine.
BZ: But at that time you already got married?
LR: Yes, I, I got married early, in, after my bachelor`s, basically in, in 1959.
BZ: So, uh, uh, but after you go back, basically, continue your work as a PhD student, you know, there's not much big changes there or there's, a, big changes?
LR: No, no, no; big changes. I moved out towards Lincoln, uh, and, and worked at Lincoln, and that continued until I moved to Washington (BZ: Okay) to do the ARPA activity, and then, in Washington, I did Telnet as well.
BZ: Right, so, what do you, eventually, so, how did it happen there? Who approached you for the ARPA and then, eventually, you moved to Washington D C and work that out?
LR: Well, I mean, the reason I started was because of Licklider’s comment that it was important, yeah, I mean, he was a peripheral scientist; he didn't know how to do it but he thought it was important to connect the computers, somehow. And I, immediately, saw how that would have saved me all this work with Minsky moving in pictures, and everybody would have that kind of problem, and all of this software that was being developed was useless, if nobody around else could use it. So, everything done, somebody did on one computer would be, sort of, in isolated from everybody else, if they didn't, if we didn't have a network. So, I, I took that on as a task. I figured, I kept choosing the thing that would be the first, I mean, graphics was the first, and the problem with that was, I was 20 years ahead of my time, and 3D displays today and, and virtual reality or something we are now talking about. But that's very, 50 years after I did what I did. But I knew it was going to be a long time. So, it wasn't worthwhile me continuing in graphics, because I couldn't sell it. Nobody could use it, it was too expensive; the computer I had was bigger than anybody, anything, anybody else had. And, and, so, what I, what I decided was the networking would be the right field (BZ: Okay), ah, where I could do something new.
BZ: Aha, that's the program run by the military, is that right?
LR: ARPA is as, is the, as an element of the DoD [Department of Defense] run by the military, is run independently under a special program, under the director of ARPA. And, so, is not really military, and the whole program isn't really, uh, part of the military in the strict sense, but it is, it is under the DoD budget. And it does most of the research for new things. Generally, should be something that's going to help the military as the network boot and, so, we had no trouble with Congress at that point, raising… I mean I, my, my plan that went to Congress was for about 15 million to do the networking activity.
BZ: So, you got to work in Washington DC, you began to work with Bob Taylor and some other…
LR: Bob Taylor was there in the office. He was the behavioral scientist who took over after Ivan. It, it wasn't appropriate for running the computer science programs, so, as soon as I came, I took over all the computer science programs (BZ: Okay) and Bob started working with the military in various areas and, and computing. And, so, he, he did his own thing in launch pad; he ran over to Vietnam and worked with them, some in other places.
BZ: Okay, uh, so what year you began to work at ARPA?
LR: Sixty-seven, I mean, I got there in late 1966 and started, but it was actually 1967 (BZ: Started) and I started, and in, in mid-1967, we had this meeting which has been reported at, uh, of all of the principal investigators of the ARPA program, which was 30 year sort of contracts and people that, principal investigators, came to a meeting. And, and we, I told them all that we were going to build this network and they were only part of it, and they needed to share the computer which they didn't like. But they, they went along with it. And Wes [Wesley Clark] suggested at that time, uh, that we separate one of the functions that we had done the initial activities, uh, uh, on my test with, we broke programs everything into the computer. So, we didn't have an ancillary computer to do the networking. But he suggested separate that out, so that it's common and in a mini computer. And mini computers were just appearing (BZ: Right), so, we did that, and, you know, I also got, um, met with people from Donald Davies group in England at a conference in 1967, and they mentioned that there were 50 kilobit lines. I didn't know about them, because I had never heard of them; 9.6 is the highest speed modem that anybody had. But, back in the backroom, unannounced, the AT&T had a huge box that, was, took seven telephone lines or nine, a lot of telephone lines to do a 50 kilobit thing. But under Telpac, it was economic for me in the government, so, I could go build the network with 50 kilobit lines, which is a far better approach than using a lot of 9.6 kilobit lines. But, anyway, he also gave me the word packet which was from the English (BZ: Yes, there is, ideas…) and I, I said that's great, it's, I needed a word; message, p…, you know, message part is not good, you know (chuckles).
BZ: Nineteen sixty-nine seems to be a quite important year, and in your, also, career. You became, uh, IPTO [Information Processing Techniques Office] director on that year, right？
LR: Yes, but that didn't change anything for me, really. Because I was running, operating as director all the time (BZ: Oh) I mean, Bob was behavioral scientist and he did his thing, separately. And I ran the programs and chose a different, uh, and ran everything and worked on the network at night.
BZ: So, you got to encourage, you know, ARPA. Said (sic), let`s build network, because we need that. So, there is a little bit debate, said (sic), um, we built this network because we want to survive a nuclear war (LR: That was Paul Baran`s content, it wasn't mine). I, yeah, so, I’m, sometimes, just to, try to understand, and, you know, you're, you pioneers and geniuses work together, what you're trying to do. You still treat the network like a toy? So, we can, you know, make better communication between scientists? Or, you are really envision or foresee (sic)e, oh, my gosh, someday, internet will be, like everybody could use?
LR: Oh, yes, I remember, in fact, if you look at one of my papers, um, uh, is, is a hand-held terminal with, with that, you could use on the network because packet radio was one of the things I designed. And, um, I did a lot of work on the theory of packet radio, and seminal paper here is probably one of the most quoted papers, because everybody today is trying to redesign and do packet-to-packet air wave, you know, radio activity, and slotted ALOHA was the original technique that I developed, uh, and, and, did so. Basically, I was very involved in the packet radio, and I knew that we would get to cell phones, I mean, that was, that was actually, (unintelligible) the picture of one of my papers.
BZ: What year was that paper published?
LR: Oh, I love to look it up because it was very early, it was early. it was (BZ: Wow); yea, it was in the 1970s, sometimes I forget.
BZ: So, what`s your thinking source behind the, the, you know, the ARPANET, or…
LR: Well, maybe my thinking was, it was going to connect all the computers, and it was going to replace voice, and it was going to replace all of the other data communication mechanisms around, because it was going to be online all the time, uh, and, and, high capacity, and immediately there. And it was so much better than what we had. Email showed that-- when it was so much dramatically better than what, what was available, uh, with the mail or with the telephone (BZ: Telephone, telegraph; so, you are really..), um, you know that, economics, and I found a lot of that too, and showed how dramatically cheaper it was. It was so cheap that it's free, basically (BZ: Okay). I mean, you don't get charged per message, or, it`s not worth it.
BZ: Oh, okay. So, I heard that, this is what I say, is, there’s a preconditions, for, this happened to everybody, is, the cost of this hardware should be, keep dropping. Do you foresee, did you also foresee that? And, you know, your computer is so expensive…
LR: Yes, in fact, that’s, that`s sort of an annoyance, because it's called Moore’s Law (BZ: Yeah). I wrote it first, um, I wrote a paper when I first got to the Pentagon. And I saw that they were releasing computers over long periods of time. So, I, I took the cost of every computer built, to, today and speed and I plotted that on a graph and I looked at this trend which was perfectly good trend, and that trend turned out to be, uh, 18 months, uh, 18 months to doubling for a computer cost, for computation. [Gordon] Moore looked at it from the semiconductor point of , and a year or so later, he published his paper saying it's a factor of two. Later, he adjusted it to 18 months. I come up with 18 months to begin with, just by looking at the actual, actual costs of computers, and I published that inside the Pentagon for these people who are leasing their computers over much too long a period in showing them that they could cut it way shorter and save a lot of money by getting the next generation computer. Um, anyway, so that, that was one of the papers I wrote. I was very convinced about Moore`s or my law at that moment. Uh, uh, what the economics (sic) were doing with electronics. So, I knew it was going to be cheaper and cheaper over time, so that wasn't really the problem; everybody would have a computer. I didn't know how many there would be to begin with. I, I didn't calculate numbers, but I knew that we could get to the hand-held devices.
BZ: That's quite amazing, that's; and, very visionary.
LR: So, in 19…, that, that's sitting right behind you on the floor, no, on the floor (BZ: On the floor), and the big picture, Ericsson Award (BZ: Aha), that one. It, basically, is the award of the, uh, uh, the, uh, from Sweden by the Nobel, uh, by the Swedish National Academy, it's the same, the same people who do the Nobel Award, and it was made by the same decisions. But Nobel`s Will didn't allow math or computer science, so, uh, Ericsson founded that award, but it went to, through the same process. and the king awarded Kleinrock and I this award back in, uh, 1981, I think it was. And it was, it was, there that I gave a speech that said, well, the network is perfectly good for voice as well as for data, uh, and we've already shown that, we took that very early, uh, in the internet activities in back in the 1960s. I mean, Danny Cohen had put together a voice at Harvard and made it work over the network. Uh, we knew that, that would work fine, and I knew that it worked from the start, because the, the speed was there. But, but, the, um, but the question was what was going to happen to the voice network over time? Well, it was already cheaper to do it on the internet back when I, when I did, when I made the speech. But I said it would gonna take 20 years for that to become reality because of the speed of change of the, of an industry, like the telephone network. and it's changed, in about 20 years, but about, uh, 2001 or so, we got to where a lot of the voice started being on the internet as opposed to, um, circuit-switched line even though we have, they may have offered circuit service, circuit-switch services, a lot of it was transported over the internet by then, and, of course, more and more was transported over the internet totally with cell phones, and, and, that was the change that put voice; you didn't normally talk to your computer but, your cell phone-- you did.
BZ: Right, right. And also our office song is already, and, put on the internet, too, you know…
LR: Yeah, that's all happening now. So, I mean, basically, everything is Voice-over-IP (VoIP) now. But that was, I predicted that would take 20 years for the telephone companies to change, and may move that, and it did; took that long. But I knew it was going to happen over time, and I knew what the time frames were because of the speed of everything.
BZ: Here's like, is, my, my question: Why is industry don't want to change so quickly while the technology is ready, and the demand is, you know, there, and, and also in some way, you know, you have to change eventually. Why the industry works, sort of work, behind the scientist?
LR: Well, first of all AT&T, when I was ready to leave ARPA, um, uh, in '60… uh, you know, in 1973, uh, I said: Okay, I’ve, I’ve got the network running; I’ve done everything I need to do here; I’ve spent my 6 years, so, I’m gonna, I’m going to, uh, go start Telenet; that’s a new career. Uh, uh, but before I decided to do that, I knew that from ARPA`s point of view, they shouldn`t keep running the network. It was not a research project anymore. It was an operating activity; it was, it was, uh, the operating network. So, um, what, what I did was I went to AT&T, and I said: I will give you all the equipment and you own all the lines anyway; we'll start paying and lease them, and you can charge for the network to everybody, including us, and build from there, and BBN [Bolt, Beranek and Newman] is building the equipment. You can expand it as much as you want.
They had a big committee study, this for months, and, finally, decided: No. It was not in their interest to be involved in packet-switching; um, it took them about five more years to change their mind, uh, and even when they tried to do it over and over, they couldn't do it, because their, their senior engineers just weren't thinking that way; they were thinking circuit-switching. They, they couldn't believe, you, the packets wouldn't fall on the floor, uh, they just, they just co