【OHI】Robert Kahn interview selection

What is the language computers used to talk to each other? And they came up with an approach called NCP, which worked really fine for the ARPANET, but had certain deficiencies. So, it didn't really extend to an internet environment, which is where the TCP/IP development comes from. Uh, and, so, this group came up with the protocols. And then all these other institutions designated people, mainly DARPA-funded, to actually figure out how to get their computers connected.

Robert Kahn

Guide to science 

In fact, I think, and you asked me the question about what kicked off my real interest in science. But I would say, what I like to say is he kicked off my interest in research. Uh, in the last year of my, uh, time at City College, he said, you know, “we have a way of, of having, kind of, special seminar programs for certain students, and faculty has to decide to do it”. So, he set up a separate program for me in that last year, as I recall, to work on certain advanced topics. And it was, it was a challenge, you know. Not, this was not curriculum-type stuff. It was, let's pick an area, always thought about before, and see how we like, you know, enter into it. I don't even remember the set of topics we dealt with, but I do recall, it really stoked a real interest, on my part, in research. And, hey you take a problem that nobody knows how to think about and delve into it. He, sort of, nurtured that for a bit of time.

Linked the Information Highway 

When I was at BB&N, we had won the contract based on this proposal to build the packet switches for the net, so DARPA would order the lines to hook these switches up. The switches determined how the network worked. Our job was not to connect the computers to it, but make it possible for computers to connect to that network so that you can communicate with each other. 

And so that task was given to all of the institutions that DARPA was supporting. And they funded some people to work on the protocols. In fact, that's how I got to meet Vint Cerf for the first time. He was working in Len Kleinrock’s lab at the time. And he and a colleague of his, name Steve Crocker and several others ended up being tasked by DARPA to work on the protocols, uh, which was in really a new area at the time. 

What is the language computers used to talk to each other? And they came up with an approach called NCP, which worked really fine for the ARPANET, but had certain deficiencies. So, it didn't really extend to an internet environment, which is where the TCP/IP development comes from. Uh, and, so, this group came up with the protocols. And then all these other institutions designated people, mainly DARPA-funded, to actually figure out how to get their computers connected. And the, the whole effort on this demonstration was to essentially put an ARPANET node in-- this was at the Washington Hilton hotel. In the ballroom, we built a false floor, did all the wiring. 

We've got connections into the other links in the, in the ARPANET. And so the, the computers were all around the country. They're already on the ARPANET. So we did was provide an ability for people to access those computers using the ARPANET from the ballroom at the Hilton, we managed to get 40different terminal devices of different sorts, ranging from teletypes to graphic displays to various things. So, people could actually use these facilities.

 If you've heard of the fellow named Bob Metcalfe; he started a company called 3Com on the West Coast, and it's probably best-known for creating, along with the, the box, something called the ethernet. He actually worked out all the scenarios. That's the task I gave to him--figure out what people would do. If you sit down at this terminal, what do they do? And so, yeah, play all kinds of scenarios, try this one try this one try this one, running on different machines. And so people could walk their way through. It's like, you know, going through an amusement parkwe got to stop here, then stop here, then stop here. You can stop at this terminal. Here's a book you stop at this terminal. Here is a book. 

And, so, and people were manning it, and some were demonstrating their own capabilities. And, so, people gotta chance to experience networking first hand. That's what that demonstration was about. And it was, it took about a year to put, put in place; make it happen. And, we had to work through all the union rules for who could do what in the hotel. And, you know, how could you describe on displays what was going on? How do we check it out? Make sure it works, all that stuff. But it was a great effort. It was, uh, really commissioned by DARPA. But it was Larry Roberts who's, who, basically, asked that we do it there. 

I had proposed to him that we do it at another event in the spring, because I thought this would be a way of coalescing, getting everybody to get things to work, kind of a forcing function to getting the ARPANET actually work. Because even though people talk about the ARPANET, you know, first four nodes being installed between September and December of 1969, . the fact of the matter is, in 1969, there were almost no machines connected, and the ones that were the best that they could do it might be too proceed the packet say I got it. Um, and then throw it away because there were no protocols to use. There were no applications running; nothing. So, the question was: how do we take this basic piece of infrastructure and get all these machines to now work properly over the net. And, so, that was what this task was about, because by 1971, when I decided, that's what I would try and spend the next year doing with the DARPA’s support. Uh, we haden't made a whole lot of progress. There was still a lot of machines that were not yet connected or didn't work properly. It could only do one out of 15things they were supposed to be able to do. And, so, we spent that year and we, basically, got all those machines to, to work together.

TCP/IP& a coin

We were at his office at Stanford and we have the discussion. I think it was Vint idea. Let's toss, let's toss a coin, which was interesting on his part, because, normally, people put things down alphabetically. And, so, alphabetically, Cerf would come before Kahn. Uh, so, he could have said, you know, let's do it alphabetically. 

But he didn't. He said, let's toss a coin. So we did, we tossed the coin, and he won the coin toss. In reality, uh, if you look at the early papers that were written about the protocol, before it was known as TCP/IP, in fact, we called it just TCP in the paper, because all the IP stuff was bundled inside of it. We didn't say, let's separate out TCP as a separate module and protocol. I mean, it, it worked essentially the same way, functionally, in terms of IP for routing and the like, but it was one module as opposed to two modules. 

So, um, so we, we tossed the coin for that. Uh, all the early papers that were written referred to it is as the Kahn-Cerf protocol. I don't know why they did that, but maybe because I was running the office and he reported to me or I don't, I don't really know what. But if you go look at any of the archives, if you can find anything that deals with that, you will see the early references wear to the Kahn-Cerf protocol. It's okay to me, would be either way, because this really was a joint effort on our part to actually create it. Um, Vint played a much more role in, actually. The boots-on-the-ground kind of thing. Uh, I was much more into, from there on, in administering it, masterminding it, orchestrating it. And, of course, I invited him to come join us at DARPA. And, so, he took over those programs from me in 1976 that he ran them for a while reporting to me in a higher level. 

And then he left DARPA in 1982 to go to MCI I took over the things that he had been doing and trained Barry Leiner to take them over, and I transferred a lot of those to him later on.