On October 29, 1969 at UCLA, Professor Lenonard Kleinrock and his team sent the first message over a network of computers that would evolve into the internet. The world’s leading technology experts and visionaries gathered at UCLA on October 29, 2019 to honor the significance of this moment and discuss the current state and future aspirations of our connected world.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
1
President Eisenhower Created the Advanced Research Project Agency
Speaker 1: Today, a new moon is in the sky. A 23-inch metal sphere placed in orbit by a Russian rocket.
Leonard Kleinrock: In 1957, Sputnik went up. It caused a great distress for this country. We were now behind the Soviets in technology. President Eisenhower created the Advanced Research Project Agency.
Steven Lukasik: ARPA was created in 1958. The guiding principle was ARPA did things that no one else was doing. And therefore, it better do it.
Bob Taylor: And their first research programs were space programs because Sputnik had been a space program.
Steven Lukasik: The ARPA director got cross-wired with the President over whether ARPA should be the launcher of all satellites for the United States. People had thought we better not let too much of the space technology stay in Defense. So NASA was established.
Bob Taylor: And these space programs were transferred from ARPA to NASA which left ARPA with room to start research in other areas. And computer research was one of the areas they opened up.
Steve Crocker: ARPA as an agency is divided up into offices. Each office has a director and a handful of program managers and that was it. So an extremely lightweight, nearly flat structure. The Information Processing Techniques Office was where all of this funding for advanced computer science was coming from.
J.C.R. “Lick” Licklider: The thing that makes the computer communication network special is that it puts the workers, that would be the team members who are geographically distributed, in touch not only with one another but with the information base with which they work all the time.
Charles Herzfeld: Licklider came and gave a seminar at the Pentagon, even talked about the new way of computing. By that time, everybody had seen the ORD batch process. You give a program your problem, this goes away in programs and finally comes back with a stack of paper in which it is alleged that the answer is hidden. That’s a miserable way to do research, but Licklider said, “This is all wrong, you need direct interaction by the individual scientists and the real machine and you don’t have to be in the same institution at the same place. You can do it all by some magic distant medium which we call the net.”
Bob Taylor: A year or two later when ARPA had created this computer research program, they invited Lick to come and be its first director.
Leonard Kleinrock: Licklider basically became the head of the computer side of this research effort.
Charles Herzfeld: I think Licklider is the smartest man I’ve ever met.
Alan Kay: Because he wasn’t a technologist, because he hooked up to this big idea, he was a big… He was big. Lick was big. He had a vision as to what a network might do, give people connectivity and they’ll do some wonderful things. He had no idea how to do it.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
2
Why Did You Pick October 29th, 1969?
JOHN MARKOFF: He left UC Berkeley after taking all of their computer science courses. So Charlie, why don’t we start with you? How did you end up running software for this machine?
CHARLIE KLEIN: Well, I had gone to UCLA as an undergrad and I’d already learned to do some programming from my dad so I ended up…
[chuckle]
JOHN MARKOFF: Go ahead.
CHARLIE KLEIN: So I ended up having to take a programming class and went to the department chair and said, “Do I really need to take this?” And he said, “No, you don’t need to. Why don’t you work on my research project?” So I was working for Gerry Estrin on the research, and then, Len got this project to work on the ARPANET. And most of us switched over to Len’s project and I was working on trying to keep our computer system working and working on the operating system and then when the IMP came, I had to write some software to make it work.
JOHN MARKOFF: And you were building an operating system, mostly though, right? For something called the Sigma 7.
CHARLIE KLEIN: Right. Right.
JOHN MARKOFF: Okay. And how did you end up being on campus at 10:30 tonight, 50 years ago?
CHARLIE KLEIN: I enjoyed programming.
[chuckle]
CHARLIE KLEIN: I sort of described it as when you got a programming working, it’s like moving up the next level in a video game. You got a little excitement. So I would often stay late at night, sometimes I stayed all night and programmed all night, then went to my classes in the morning.
JOHN MARKOFF: Okay, so the first connection that you had with Menlo Park was not seen as a big deal by you?
CHARLIE KLEIN: Not to me.
[chuckle]
JOHN MARKOFF: And why did you pick October 29th 1969 to do it on?
CHARLIE KLEIN: I’m not sure there was any specific reason for that date, we were ready to try it, Bill was ready to try it and we started trying it.
[laughter]
JOHN MARKOFF: Bill, how did you come to be on the other end of that communication?
BILL DUVALL: How did I come to the…
JOHN MARKOFF: To be on the other end of that communication?
BILL DUVALL: Well, basically, I was working with, at SRI with Doug Engelbart and his group, and that’s where a lot of the paradigm that we see today on the internet came from the idea of using a computer, which up to that point, had been used for basically databases for bank and for numerical processing. The idea of using it as a page-oriented, with a mouse, information system, and in fact, that’s where the mouse… He invented the mouse as part of that whole effort. But part of the concept was that there would be basically a lot of workstations… He called them knowledge centers, basically, and these would all have these types of terminals and they’d be connected via network. And so, when the idea of the… And it was being funded by ARPA, so when the idea of a network came along, it was a very natural step to say, “We’ll be one of their first nodes.”
JOHN MARKOFF: You and Charlie hadn’t physically met?
BILL DUVALL: No, in fact, we never physically met for… Until I think 10 years ago or 15 years ago, so.
CHARLIE KLEIN: Yeah. [laughter]
JOHN MARKOFF: But you did regularly talk on the phone?
BILL DUVALL: We did talk on the phone. The whole… When the IMP was delivered, which was, I think the beginning of October at SRI, along with the IMP came a deadline and the deadline was October 31st to basically have a login from one computer to another over the ARPANET at that point. And so, we worked furiously to try and get all the software ready do that, and in order to get… We knew that everything was… There were a lot of things that we were putting together. It was a complicated system, a lot of pieces, and so we decided to start testing early and that’s why… It was October 29th, to see if we could beat the deadline, and so we scheduled a test for 29th, it was at 10 o’clock at night because there were other people that used the system during the day and we knew that it was fragile to say the least, this was pre-alpha stuff.
[laughter]
JOHN MARKOFF: But also, this wasn’t your primary job at the Augmentation Research Center, this was something you were sort of called into the last moment to do?
BILL DUVALL: That’s right, yeah. There was… A person had been contracted to do the interface and when the IMP was delivered, doing a review of the state of software was obvious that it wasn’t gonna get there and at that time, I was the only person in the group that both understood the protocols and could work in the SDS 940 Time-Sharing System. So I sort of, I was volunteered, let’s put it that way.
[laughter]
JOHN MARKOFF: So, this is the “Watson come here quick” moment of the internet and nobody really knew that that was the case for maybe a decade or two, is that right?
*Internet50 content is owned by UCLA Samueli School Of Engineering.
3
When the Idea of a Network – We’ll Be One of their First Nodes
JOHN MARKOFF: Bill, how did you come to be on the other end of that communication?
BILL DUVALL: How did I come to the…
JOHN MARKOFF: To be on the other end of that communication?
BILL DUVALL: Well, basically, I was working with, at SRI with Doug Engelbart and his group, and that’s where a lot of the paradigm that we see today on the internet came from the idea of using a computer, which up to that point, had been used for basically databases for bank and for numerical processing. The idea of using it as a page-oriented, with a mouse, information system, and in fact, that’s where the mouse… He invented the mouse as part of that whole effort. But part of the concept was that there would be basically a lot of workstations… He called them knowledge centers, basically, and these would all have these types of terminals and they’d be connected via network. And so, when the idea of the… And it was being funded by ARPA, so when the idea of a network came along, it was a very natural step to say, “We’ll be one of their first nodes.”
08:11 JOHN MARKOFF: You and Charlie hadn’t physically met?
08:13 BILL DUVALL: No, in fact, we never physically met for… Until I think 10 years ago or 15 years ago, so.
08:19 CHARLEY KLINE: Yeah. [laughter]
08:20 JOHN MARKOFF: But you did regularly talk on the phone?
08:22 BILL DUVALL: We did talk on the phone. The whole… When the IMP was delivered, which was, I think the beginning of October at SRI, along with the IMP came a deadline and the deadline was October 31st to basically have a login from one computer to another over the ARPANET at that point. And so, we worked furiously to try and get all the software ready do that, and in order to get… We knew that everything was… There were a lot of things that we were putting together. It was a complicated system, a lot of pieces, and so we decided to start testing early and that’s why… It was October 29th, to see if we could beat the deadline, and so we scheduled a test for 29th, it was at 10 o’clock at night because there were other people that used the system during the day and we knew that it was fragile to say the least, this was pre-alpha stuff.
[laughter]
09:16 JOHN MARKOFF: But also, this wasn’t your primary job at the Augmentation Research Center, this was something you were sort of called into the last moment to do?
09:23 BILL DUVALL: That’s right, yeah. There was… A person had been contracted to do the interface and when the IMP was delivered, doing a review of the state of software was obvious that it wasn’t gonna get there and at that time, I was the only person in the group that both understood the protocols and could work in the SDS 940 Time-Sharing System. So I sort of, I was volunteered, let’s put it that way.
[laughter]
09:49 JOHN MARKOFF: So, this is the “Watson come here quick” moment of the internet and nobody really knew that that was the case for maybe a decade or two, is that right?
*Internet50 content is owned by UCLA Samueli School Of Engineering.
4
Request for Comment and the Basic Plan for the ARPANET
JOHN MARKOFF: Well, so then, let’s talk a little bit about how this came to be. Steve, there was this process called Request for Comment and I think that that… If you could explain that and how that came to be, ’cause that was really where the architecture for the system came from.
Steve Crocker: The basic plan for the ARPANET had been laid out by the adults, we were the kids. And so, the IMP was designed by Bolt Beranek and Newman, and the long lines, the telephones that were procured from AT&T. And then what seemed, in a way, an after thought, was okay, “I gotta get some software running on the host to talk to each other.” There was a meeting called at Santa Barbara, UCSB and Elmer Shapiro from the SRI group chaired the meeting, Vint and I drove up, we were… Did you make the choice to send us up there? Somebody said we should go…
Leonard Kleinrock: I wanted to go the cheapest way, you kept busting my budget.
[laughter]
STEVE CROCKER: That came later, the trip to Santa Barbara was cheap, the expenses grew after that. So, Vint and I… This was August ’68, so this was more than a year before the time that we’re talking about. The contract for the IMPs had not yet been let but it was in process, we met counterparts from the other groups. Bill, were you at that meeting? Jeff Rulifson was there, I remember.
Bill Duvall: Yeah, I was there. That was Jeff Rulifson, [12:31] ____ Elmer Shapiro and…
STEVE CROCKER: And it was a pretty interesting meeting in that there was not much of an agenda, we all came with… To find out what was going on and it was sort of like a cocktail party engagement, where you instantly click with certain people, that you can tell you’re on the same wavelength. And we could see that this communication was going to be more interesting than just logging into a remote machine, or sending a file, those were obvious… We obviously wanted to be able to do that, but we could see a bigger picture and we kind of self-organized after that and decided that we would keep talking, and in the process to keep talking, we said, “Well, we should visit each other’s laboratories,” which is were Len’s comment about busting his travel budget came from. And we understood the irony right at the very instant that we said that, that this network was supposed to make it possible to collaborate at a distance, without having to travel. And the first thing we did was lay out a whole travel schedule, so we could visit each other’s laboratories.
[laughter]
STEVE CROCKER: Those meetings, we engaged in large and small topics. The specification for the fine grain communication had not yet been specified ’cause BBN was not yet working on this and they hadn’t published their interface specifications. So we looked at the bigger picture what kinds of things would you wanna do? We sketched out a number of interesting ideas and we did this just sitting around the table in these occasional meetings every six weeks or eight weeks, and then in the spring of 1969, so about six months before the date we’re talking about here, we said, “Well, we should start writing down some of these ideas after all we’re supposed to be academics and we’re supposed to be doing research, we should write something down.” So we dealt out assignments to each of us, “You write this, you write this, I’ll write this” and then I casually offered that I would organize the notes, which I didn’t think about it at the moment, but over the next couple of weeks, every time I sat down to write what should have been a trivial administrative clerical note, I found myself balking big time, I had trepidation and I realized what it was, it was that the act of writing things down might make it look like these were official important authorized and I was very fearful that some adult was gonna come from the East, I didn’t know whether it would be from Boston or from Washington, but “Who are you guys and who gave you authority to do any of this?”
JOHN MARKOFF: Let me quote from RFC 1, you wrote “very little of what is here is firm and reactions are expected.”
STEVE CROCKER: That’s right, that’s right. So late one evening when I was, I was determined I had to write something down I said, “Look, these things have no status, no authority, write anything you want, write as little as you have in mind, you can write questions without answers and so forth.” Bill had suggested the term “request for comments”, I hadn’t remembered that until recently but it stuck in my mind and I said, just as a matter of form, we will make every one of these notes called a request for comments. I figured it was a temporary hack last maybe six months until there was formal documentation of this network. And the other rule that I put down was that you had to write it before I give you a number ’cause I didn’t want a lot of holes in the sequence. And so that seemed to work pretty well except that it didn’t stop, it just kept going and going and going and when I was asked… We did an index of the first 100 and then I was asked to write something for the first 1000 I thought, oh this is a sorcerer’s apprentice kind of situation where you can’t turn it off.
Vint Cerf: They’re about 8800 RFC.
JOHN MARKOFF: Today.
STEVE CROCKER: Well, and they’re not the same because in the early days we had no other way of communicating. We sent these out by snail mail, by US postage, we had a mailing list and in fact, the mailing list was maintained as one of these RFCs and every time we added somebody to the mailing list, it was another RFCs, another RFC.
JOHN MARKOFF: Bill remembered that you guys flipped the coin to see who would write one and two, is that, your memory?
STEVE CROCKER: I lost, I got RFC 1.
[laughter]
Bill Duvall: I remember it the other way, you won and got RFC 1.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
5
ARPANET and how to make it flexible Enough to Accommodate the Future
JOHN MARKOFF: What was remarkable to me about RFC 1 is that it was about two experiments. Did you guys know much about the online system when you actually set that up?
STEVE CROCKER: We had visited this process from August ’68 to spring of ’69, we had now visited each of the laboratories. What was remarkable particularly in retrospect, was the SRI laboratory, Engelbart’s lab, had… They had invented the mouse and ordinary people didn’t see it for another 30 years or something like that, approximately and a graphics interface with hyperlinks and structured text and everything. So the future was sitting right there and better yet, it wasn’t just you could take a tour and see it in laboratory, that group was using it every day. That was part of their basic thing and so it just had to wait ’til it became commercially viable. But we were living, at least, thinking in a mode of, this is the way the world is gonna be and we were gonna lace it together with this ARPANET and then how could we make it flexible enough so that it can accommodate the future technologies that were coming along?
JOHN MARKOFF: I wanna ask a little bit about the culture of those laboratories at the time that you guys were doing this work. You guys both started early, Steve and Vint, you started as high school students spending regular time on the UCLA campuses, is that right?
STEVE CROCKER: Oh yes.
JOHN MARKOFF: Breaking in even on occasion.
STEVE CROCKER: I was afraid you were going there.
VINT CERF: Wait a minute, Steve got permission from Mike Melkenoff who chairman of the department at the time to use a Bendix G-15 paper tape-based machine at UCLA and I think eventually we got to use the 7090, so we had permission. But one evening we showed up or maybe it was a weekend, we showed up and the door was locked. And so what happened then, Steve?
[laughter]
STEVE CROCKER: Statute of limitations has passed. Vint and I had been playing with some silly equations and decided we would try to explore them numerically, at least… It sounds better now than whatever I was able to say then. So this Bendix G-15 was in a building they called Engineering 1 which I believe is now gone but I did have permission to use it and we arranged to come over from Van Nuys High School in San Fernando Valley, it took a little less time to do than it would take today I think.
LEONARD KLEINROCK: You know the dean is in the audience.
STEVE CROCKER: Tell everybody. So we came over on a Saturday, and the building was locked, and I was crestfallen. Two of us are standing there staring at the door, and Vint notices that the second floor window is open and so one of these crank things. And I’m thinking, we’re not really going to… Next thing I know, he’s on my shoulders.
VINT CERF: I always have, that was the beginning.
STEVE CROCKER: He goes through the window, he comes around to the main door, pushes the bar open, and we taped the door so that we could get in and out easily during the day to go buy some food at lunchtime or whatever. Nobody else was around. The interior doors were all open in the offices. This was long before student demonstrations and all other kinds of things that have beset us. We worked all day and we cleaned up afterwards and we went home. That was spring of ’61. Eleven years later in ’72, I’m working for the Defense Department, I’ve got a top-secret clearance, and the Watergate burglars break in, and they taped the door. And one of their cohorts who’s in the hotel across the street is on lookout, but he doesn’t recognize the threat because the security guard who found it was in plain clothes, and the shiver that went down my spine was something.
[laughter]
JOHN MARKOFF: So… Go ahead, you can tell your side of the story.
VINT CERF: I was at the White House in 1999. Hillary Clinton was doing her Millennium Evenings, Honor the Past and Imagine the Future, and she told that story. So the question is, how the hell did she find out about that?
*Internet50 content is owned by UCLA Samueli School Of Engineering.
6
Was it a Hacker Culture?
JOHN MARKOFF: So Steven Levy has written a book called Hackers, you know that… The term “hacker” meant one thing before it meant something else, and when you guys were first computing in the ’60s, a hacker was someone who was really obsessed with computing. And there was a laboratory at MIT, the iLab, where this culture grew up. Would you call the culture you guys created, both at Menlo Park and LA, similar? Was it a hacker culture in that first sense of shared information?
VINT CERF: Well, it was MIT that adopted that term. You were there.
LEONARD KLEINROCK: I was there, but there was an interesting story. These guys again are guilty. There was some guy on campus that committed some terrible break-in to a computer system. Vint decided to hire him as a programmer. Smart guy, you fired him a few months later, right? Didn’t work out.
STEVE CROCKER: It was a question as to whether he could learn to do something other than breaking into systems, particularly if we said, “Look, it’s not interesting, we’ll give you the keys to everything and see if you can do something constructive,” and…
VINT CERF: So hack was a really cool piece of software. The hacker was somebody who produced really clever code, and that was an honorable title for many years.
JOHN MARKOFF: Bill, was the culture the same at Menlo Park?
BILL DUVALL: No, I wouldn’t say it was at all. The culture in Doug Engelbart’s group, the one that I was in, was basically almost self-contained, and people were very focused on this one vision that we were working on. And, although there was some connection to the outside and interchange, it was very little. The group, in and of itself, was a little bit odd, and for SRI, it was generally disliked, I guess I would say. It was certainly not understood. They did things like… We’d work at all hours and we’d have… Friday afternoons would be for wine and beer, and… This was very un-SRI-ish. And one of the things I remember is when I was interviewing to join the group, the last interview had to be with the head of the engineering department, and he asked a few questions, “Why do you wanna do this now?” I’m saying, “What they’re doing is the future, I want to be with that.” And after a while, he thought about it and he looked over his desk at me and he said, “Son, you don’t think what they’re doing up there is science, do you?”
LEONARD KLEINROCK: John, you know you’re talking about the culture for the period, and there is a very strong component that mattered and helped create what we have today, and it’s the way in which the funding took place. I think you’re familiar with the story, the way ARPA funded the various centers, the universities and… They would go to a great researcher there and say, “Here’s a pile of money. Go do something great. Shoot for the moon. Failure’s okay, go high, keep going, you have the money for a long time and we’re not going to watch you.” Now, what else could you ask for? It was an environment that just generated… It was a golden era. And when that came to a principal investigator, passed it off to these guys. Same idea. They complained we didn’t supervise them. I said, “Go do it. Eventually, if you want something, come back to us, but… ” And they ran with it. You can just hear the stories, the way they organized themselves and created this community across many universities, was a very important component that gave a lift to this whole project.
JOHN MARKOFF: Well, and how quickly did that end? And was it the Proxmire Amendment that ended that era of…
LEONARD KLEINROCK: Not totally, but it hurt a lot. They said, basically, it has to have a military application, has to be competitive. You go to a guy, here’s the money, that’s not competitive, it’s not peer-to-peer, but it worked.
VINT CERF: Well, on the other hand, ARPA has some pretty smart people, and so they created this thing called a Broad Area Announcement which said, “This is the area in which we would like to explore. Would you like to submit proposals for that?” I think we managed to induce the right people to submit the right proposals, and the longevity of support from both NSF and ARPA is countable in decades, and so that’s still true today.
LEONARD KLEINROCK: It’s true to… But your point about having really good people at the government deciding where the money should go, people like these two here. They were at ARPA. That level of capability looking out and finding the projects to support. That whole thing worked so well.
JOHN MARKOFF: I wanna ask about the design of the network, but before I go there, you were… A lot has been made of the first crash, and I wanna ask a different question about the first crash. When you type G, there was what is known today as a buffer overflow error. And the machine crashed, and you had to fix that problem, and you started over again. What struck me when I learned about for the first time, is that buffer overflow errors have continued to plague the design of computer systems ever since. And in fact, you might say they’re one of the real vulnerabilities in the design of modern computing. What I don’t understand is when Robert Tappan Morris for example, brought the internet to its knees in 1988, the then young internet, he used a buffer overflow there. Why didn’t the designers…
LEONARD KLEINROCK: Because…
BILL DUVALL: I can answer that, it’s because they didn’t talk to me.
JOHN MARKOFF: Walk us through that.
LEONARD KLEINROCK: It’s stupid mistakes.
BILL DUVALL: There are two answers to this. One of them is that the initial specification for the first connection was that all messages would be one character in length, and because of the way that the 940 and time-sharing system worked, when you typed a character on your terminal, it didn’t appear on your terminal, it went to the computer and the computer then sent back that character. That’s called full duplex. And the 940 had what they call command recognition. So as you were logging in, as soon as it realized what you wanted to say, it would tie it back, it would send back the rest of the command. So when you type an L, you got an L, and O, you got an O, and a G, you got G-I-N. Well, that’s three characters. And there was no memory available. Memory was very, very tight in the 940. And so when I was allocating buffers and things like that for output, I allocated one character because that’s what the spec said. And three came along and there was a buffer overflow. Now, that seems to absolve me, but the problem is I wrote the spec.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
7
How Leonard Kleinrock Came to Put Together the Networking Group
JOHN MARKOFF: So, Len, tell us the story of how you came to put together the networking group that actually sort of did this on your end.
LEONARD KLEINROCK: Well, I came to UCLA having this theory in my hands, waiting for it to have it implemented. Finally, as I said, ARPA said, “Let’s build a network. So I had to get this group together. As Charlie said, he was working with Gerry Estrin. Gerry Estrin had put together a group of great programmers doing a variety of things. When we got the contract, the whole group moved over to me to jump on this really exciting project. So that’s how this group entered the picture. And they continued to run with it beautifully.
29:13 JOHN MARKOFF: So what about the design ideas? Packet switching seems to be kind of a non-obvious idea. And what was the networking theoretical world like when you were starting this process and why did you would… What attracted you to packet switching?
29:28 LEONARD KLEINROCK: So there was no world. At the time, nobody was thinking about data networking in a serious way. And all my classmates at MIT were working on information theory and coding theory which Claude Shannon had solved fundamentally. And the problems that were left over were small and hard. And I was dragged in to do a PhD. I didn’t want to. My supervisor, “You gotta do it.” So I said, “I’m gonna do it.” I wanted to do something with impact, and I wanna get the best professor I know. So I spoke to Claude Shannon and you may not know his name, but he’s a great man. He took me on, and I looked around and I was surrounded by computers and I said, “One day, they’re gonna have to talk to each other.” And the telephone network was woefully inadequate.
30:13 LEONARD KLEINROCK: So I looked at this problem. Here’s a new problem. If it can be solved, it’ll have impact. And I had an approach. I knew how to basically extract the essence of the system which had to do with what we now call the sharing economy. It’s called Airbnb. If you got a room that’s not being used, if you got a communication channel that’s not being used, let somebody else use it who needs it right now. That whole idea of dynamic resource sharing was a principle that I put into the theory. The whole idea of the packet switching, I analyzed a case where you would chop messages into fixed link things called packets, and handle them one at a time. And that had some advantages.
30:53 LEONARD KLEINROCK: So that whole thing came together. I had done the analysis, did an optimization, and then stood around and extracted the principles as to why this thing works well. Most of the students these days don’t bother to do that. They get a result and they… And “Okay, I’m done,” instead of saying, “What is it teaching me?” And we extracted some principles like, dynamic resource sharing was key, like big systems are better, wanted to study a large network, millions of nodes, instead of tens of nodes. Because then you get emergent properties you don’t see in a small network. And looking at distributed algorithm, distributed control and recognizing nobody’s in control when you have a large network, but you can’t get one node to be in control ’cause it’s too much traffic, too vulnerable, too much traffic going in and out. So distribute it. Once you distribute the control, nobody’s in control. So will that work? So you had to prove it wouldn’t collapse. And you could chop pieces, it would still work. So those are the kinds of things we extracted. Came to UCLA with that, waiting to build it, the opportunity came along as I said, and bang…
32:01 JOHN MARKOFF: And it was a DARPA-funded opportunity. Is that what you’re talking about?
32:04 LEONARD KLEINROCK: DARPA-fund… ARPA funded.
32:05 JOHN MARKOFF: ARPA-funded opportunity. Right.
32:06 LEONARD KLEINROCK: Yes.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
8
Survivability in the Event of a Nuclear War
JOHN MARKOFF: So let me jump up a level. And there’s for many years, been a debate over whether ARPA was funded or in some way back because of the design was about survivability in the event of a nuclear war. This is hotly controversial, is it wrong?
VINT CERF: Yes, it is.
LEONARD KLEINROCK: It’s an urban myth, but there’s some truth to it.
JOHN MARKOFF: Okay, Lukasik apparently made that assertion.
VINT CERF: Look, in the video that you saw, you could tell that Taylor and Roberts were trying to figure out how to share computing resources among about a dozen universities that were studying artificial intelligence and computer science for ARPA in the 1960s. They couldn’t buy a new computer for every university, every year, so they said, “We’re gonna build a network and you’re gonna have to share.” So, the whole idea in papers that Roberts and Barry Weston wrote were about resource sharing. Even though some of the early ideas of packet switching came from Paul Baran in the study that he did a RAND Corporation, and that was about building survivable networks for post-nuclear responses and post-nuclear scenario, so we should not conflate those two things. The subsequent internet design, however, did go after this question of survivability in a post-nuclear environment, and we even tested that idea because we used the packet radio system and put radios in the strategic Air Command aircraft, artificially broke up the ARPANET into pieces and then glued it back together using the TCP/IP protocols in packet radio. So there is truth to some of this, but it wasn’t specific to ARPANET as much as it was to the subsequent…
LEONARD KLEINROCK: You know, once.
JOHN MARKOFF: Steve…
STEVE CROCKER: Let me… So this is to say is that sort of the persistent myth and the question is, to what extent there’s truth there, and it’s subtle because there’s various layers. As Vint said, Bob Taylor and Larry Roberts were all focused on how do you build a system that would allow the computers to communicate, and people to communicate, share resources, and the degree of “survivability” had to do with just normal operation. Parts are going to break, lines are gonna go down and so forth. When you talk about survivability in a much more stringent sense, you’re in a different part of the design space, and it isn’t just all or nothing because the worst case, the one that attracts everybody’s attention is imagine nuclear holocaust, and nobody’s ever really focused hard on that problem. And I’ll come back to that in just a second, ’cause I spent some time with Lukasik over the past couple of years, having a discussion about that. But you also have, apropos of what Vint talked about, you have stressful situations, particularly in the military, that are more stressful than ordinary daily operation, but less stressful than the Holocaust kind of total catastrophe, and the kind of survivability and knitting the pieces back together again are important experiments there. But they still won’t take you all the way to what do you do if… In the worst case.
STEVE CROCKER: The other part of the story that I think is very important is this culture of initiative that you described, played out in multiple layers. So, you have the agency, ARPA, which later became known as DARPA, which was created after Sputnik, and inside you had a division into various offices and the office that was of concern to us was this Information Processing Techniques Office. And there was quite a bit of delegation down to the offices and to the program managers, and more importantly, even to the researchers themselves, the principal investigators, to invent the projects. And so, there was as much bottom-up kind of operation and choice of topics and so forth, as there was top-down. It was not a dictatorial thing, like we’re gonna build… This was sort of different from saying, “We’re gonna put a man on the moon,” and everything gets organized to that one objective. It was much, much broader. So, the problem that the director of the agency, there was a series of them, but Steve Lukasik was central for a long period of time, is how does he provide the support for all of these projects to go forward and justify that to the higher levels of authority in the Defense Department and Congress?
STEVE CROCKER: So, a conversation approximately like this took place. Lukasik says to Larry Roberts, “Well, could it solve the reconstitution problem?” And allegedly the answer was, “Well, I suppose.” [chuckle] And there’s no paperwork on this, except that years later, Lukasik wrote “Why I signed the cheques for the ARPANET,” and among, included in there is a little reference to nuclear survivability. But Roberts, I pressed him on that a couple of years ago, I interviewed him and he said, “Well, if I’d really been trying to solve that problem, we would have connected every IMP to four other IMPs.” So, and clearly down at the levels that we were working at, there was zero, absolutely zero attention to that. We were working on the problem that affects us all today. How do you get people to talk to each other? How do you get the computers to talk to each other? How do you accommodate the differences among all these systems and get them and so forth? And we were not at all focused on sort of the hardcore defense problem of the day.
JOHN MARKOFF: Weren’t there some… Go ahead.
VINT CERF: So, wait a minute. I’m sorry, I really feel compelled to jump in here, because there are two different things that we’re talking about. We’ve been talking about the ARPANET, that’s what this big celebration is about, on the first connection of the two hosts. But, I really must insist, Steve, that when Bob Kahn and I started to work on the internet design, it was driven by the idea that we would use these technologies for command and control. It was very driven by exactly a military requirement, and that’s why we ended up with satellites and mobile radio, as well as the original ARPANET. So, I wanna make sure that it is not misunderstood, that the motivations behind the original TCP/IP work was somehow only to do with civilian applications, because frankly it wasn’t.
STEVE CROCKER: Yeah. So, this is the area that takes some care in describing, because of course, and the project and the orientation that you’re working on is very important, but I was trying to say is that that kind of application, the command and control and reconstitution in the case that you lose various pieces, is qualitatively different from trying to imagine a post-nuclear exchange environment in which.
VINT CERF: I accept that although I will argue that we try to test at least some ideas that would be needed in order to contribute to that solution and that’s why we did the artificial break-up of the ARPANET.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
9
How Distributed Architecture Won the Day
JOHN MARKOFF: I’m also very interested in how this distributer architecture won the day. When I showed up as a young reporter in the early 1980s, IBM was pushing this idea of a Token Ring and there were many centralized ideas around. Actually, before I get to that all the computing you guys have talked about IBM and AT&T generally don’t show up in that community that you were building. They were the dominant… Why was that?
LEONARD KLEINROCK: It was a proprietary network, they were pushing it, you had an SNA network and IBM SNA network you bought IBM equipment and used their communication protocols.
JOHN MARKOFF: Wasn’t IBM and some of those large companies, weren’t they also skeptical about the idea of this project? They didn’t even bid on this, is that right?
VINT CERF: Right, we’re not skeptical about networking because IBM had SNA, Digital had DECnet, HP had DS, they wanted to network their computers but they wanted only their brands of computers to be interconnected and the defense department didn’t wanna end up trapped in one particular brand of computer for purposes of networking which is why the heterogeneous interconnection of different brands of computers which is demonstrated on the ARPANET and then subsequently expanded to allow multiple packet-switched networks should be interconnected was important to the Defense Department.
STEVE CROCKER: There are multiple times here, if you look at 1972, for example, when the public unveiling of the ARPANET a bunch of… Few good people from AT&T came and looked at it and turned up their nose and walked away. You talked about later in the late 1970s, then IBM and AT&T and others are all pushing networking of their particular flavor and then it took a while for it to settle down.
CHARLEY KLINE: If you go back to ’68 when the IMP project was being… The RFP for the IMP was… AT&T and IBM both refused to bid, they both said, “This is a waste of time, it won’t work.”
LEONARD KLEINROCK: AT&T finally launched in 1983, their great network called Net 1000, three years later they closed it down with a billion dollar loss ’cause they couldn’t do it. The technology is not trivial and they [42:13] ____ never got it going.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
10
First Demonstrations of Email
VINT CERF: Well, I think if first demonstrations of email which came in mid to late 1971 when Ray Tomlinson demonstrated that, very quickly after that.
JOHN MARKOFF: So email was not on the ARPANET in the first two years?
LEONARD KLEINROCK: That’s right, yeah.
VINT CERF: It was a development that it was foreshadowed by use of time-sharing systems for email among the participants on the time-sharing system. Tomlinson at Bolt, Beranek and Newman figured out that he could use file transfers to move messages from one machine to another, he only had to say which machine it was going to and for whom. So he used the only character on the keyboard that wasn’t already used with other operating systems, the ‘@’ sign that’s why he had user@host as the basic form for email. So as soon as email popped up and we all got excited about it, we started seeing mailing list and that’s when I realized that there was a real social element to this. The first mailing list I remember was called Sci-Fi Lovers because we used to argue who were the best science fiction writers and the next one I remember from Stanford was called Yum Yum and it was a restaurant review for the Palo Alto area. It was very clear that this technology had a social component.
JOHN MARKOFF: Do you remember HumanNets?
VINT CERF: Yes I do.
JOHN MARKOFF: Because that was what drew me to the early ARPANET because there you had technologists talking about the impact of technology which was a really interesting window into that role.
LEONARD KLEINROCK: There’s another story, I visit my graduate students, mostly theoretical guys not the software developers, and I walked into their lab one day and instead of doing researches that I wanted them to, they were busy on the machines, intensely, on news groups, restaurants, hiking, astronomy. “What are you doing? Oh my God, this is hot.”And it really caught on.
JOHN MARKOFF: It was social media. Bill.
BILL DUVALL: Just on that, one of the things that, just speaking of email, another kind of social impact was there was a project that was being done joint between Xerox and PARC and a part of Xerox in El Segundo that was being coordinated by email. This was pretty early, this was in the ’70s, mid-’70s. And the thing that became very obvious was that using email to coordinate a project shifted the power. The people that now had the power, not the people that presented well, that could talk well that were big, but they were the people that could write well, and it was just an interesting… Subtle but very interesting shift that was a direct result of basically, this distributed… The networking.
*Internet50 content is owned by UCLA Samueli School Of Engineering.
11
Why Did the Network Name Change from ARPANET to Internet
JOHN MARKOFF: So, here’s another question. How and why did the network name change from ARPANET to internet? And if they could change the name internet now, what would you call it? That’s a perfect name, it seems to me.
VINT CERF: Well, originally, we called it, the paper that Bob Kahn and I wrote said “A Protocol For Packet Network Intercommunication.” And it took too long to say that. And so, within a year or so, we used the term “internet” to refer to the multiple network thing. I don’t know what I’d call it now, probably a… No, there are some bad words that occur to me. [chuckle]
*Internet50 content is owned by UCLA Samueli School Of Engineering.
12
Moment Understood it was Going to Impact the World
JOHN MARKOFF: One last word. Steve, was there a moment where you understood it was gonna have the impact on the world that it did?
STEVE CROCKER: It was pretty evident, almost instantly to me, that if you had a computer, you would want it connected to the other computers. Now in those days, computers were big. That was a tiny computer by comparison. So, universities, and businesses, and governments had computers, personal computers didn’t exist yet. But it was evident to me that basically, that every computer if you owned it, you would wanna be connected to the network. And so, it was just a question of how long it would take for that to happen. And I thought that was real important from a utility point of view, from a practicality point of view. My head was focused on different kinds of research, much more abstract kinds of things, and so I used to sneer that this networking stuff was only socially useful, it didn’t have any real depth.
*Internet50 content is owned by UCLA Samueli School Of Engineering.