IEEE

Oral-History:Donald Davies & Derek Barber

SHARE |

From GHN

Revision as of 15:13, 12 June 2013 by Administrator1 (Talk | contribs)
Jump to: navigation, search

Contents

About Donald W. Davies and Derek L.A. Barber

Donald Davies was a Welsh computer scientist who was a co-inventor of packet switching (and originator of the term), along with Paul Baran in the US. From 1947, he worked at the National Physical Laboratory where Alan Turing was designing the ACE Computer. He became interested in data communications following a visit to the Massachusetts Institute of Technology. He first presented his ideas on packet switching at a conference in Edinburgh on 5 August 1968. In 1970, Davies helped build a packet switched network called the Mark I to serve the NPL in the UK. It was replaced with the Mark II in 1973, and remained in operation until 1986, influencing other research in the UK and Europe. Larry Roberts of the Advanced Research Projects Agency in the United States became aware of the idea, and built it into the ARPANET. Derek L.A. Barber was his deputy at NPL before becoming chair of the European Information Network and eventually moving to Logica, in the private sector. Subsequently, he headed the Alvey Directorate.

The interview, in two parts, ranges over the work done at the NPL and EIN. It covers the development of Scrapbook and Edit, two early collaborative word and data processing programs, the influence of NPL on ARPANET, international conflict over datagrams, and the evolution of packet switching.


About the Interview

Donald W. Davies and Derek L.A. Barber: An Interview Conducted by Janet Abbate, IEEE History Center, March 17, 1996

Interview #264, IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

Donald W. Davies and Derek L.A. Barber, an oral history conducted in 1996 by Janet Abbate, IEEE History Center, New Brunswick, NJ, USA.


Interview

INTERVIEW: Donald W. Davies and Derek L. A. Barber
INTERVIEWER: Janet Abbate
DATE: March 17, 1996
PLACE: Sunbury-on-Thames, England

[Note: The first part of this interview has been lost.]

Louis Pouzin

Barber:

I was out with my family, and Louis [Pouzin] said, "You must come out and join me for dinner." So there was my wife and my little daughter and son — he was not that big at the time, he was about ten. We went to this restaurant in Paris, it was thirteenth century, a lovely place. We were sitting there, and Louis has got the menu and he's sort of translating , and I remember him getting to "chou" pastries. And he was saying, "Chou," that's "cabbage," "cabbage pastry." And I'm saying, "That's all right, we call it  'chou' as well." He is going down [the menu], and I suddenly became aware of my ten-year-old son saying, "It's got it written in English on the other side!" I had to kick him under the table to stop him.

Davies:

Louis did, of course, a lot of publicity and public relations all over the world, and a lot of prolific traveling. I remember once, I was giving lectures in Brazil and I saw him there. That is a fairly unlikely place to be giving lectures, and we were traveling from Sao Paolo into Rio de Janeiro for a short holiday afterwards. And I went into the loo, and there standing next to me was Louis Pouzin!

Barber:

He certainly got around, didn't he? [Pause] So, are you happy with what we have covered, or is there anything else?

Scrapbook Use at NPL

Abbate:

I do have some other questions, if you don't mind.

Barber:

Sure.

Abbate:

Just to get for a minute to people using the network, I have seen in your paper some sort of numbers about two hundred people using Scrapbook or whatever, but I don't have a sense of who within NPL [National Physical Laboratory] this was. Was it just the computing people, or was it everyone?

Barber:

In the end, all of the divisions. For a while there was an interdivisional data processing committee, which I chaired, and we had enthusiasts, if you like, in most of the divisions, who would be trying to use computer techniques — perhaps almost in spite of their colleagues! — and they would be the people who would want a terminal on the network and so on.

Davies:

Their purpose was for some experiment they were doing. It wasn't computer technology per se.

Barber:

No, no, no. I am trying to think of the name of a chap in meteorology actually, who was doing a lot of computation to deal with whatever. He'd got some experiment there, and he was desperate to get a terminal on the network so he could do this thing attached to his equipment. So there were sort of patches of enthusiasts in other divisions who could see that they could use computing techniques brought to their bench, if you like, because of the network.

Abbate:

So you think anybody who was already interested in using computers would be likely to be interested in the network?

Barber:

Well I think so, because when you look at the alternatives — I mean, in the early days we used to go in at night to work on the ACE and single-shot through a program, and things like that. [Laughs.] There is such a difference between that and having something in your office.

Davies:

All the attractions are on-line.

Barber:

So I think it was fairly attractive. The enthusiasts certainly were keen to have it.

Davies:

Because there was no on-line service except through the network.

Barber:

That's right.

Davies:

Whereas for example when I was at MIT people were using the networks through the telephone systems all the time. Whereas at NPL that didn't happen. The first time people managed to get on-line services without coming to the neighborhood of the computer was through the network.

Abbate:

So they hadn't had any remote time sharing?

Davies:

No. Well they did, but only from public services. Right from the beginning, as soon time sharing services were set up in London, we had subscriptions to them. But that was very limited in my division. For example, we just had one terminal set aside in a room, which you could go to if you wanted at any time to try that service. It wasn't a widespread service at all. I brought a teleprinter home here one night, in order to demonstrate.

Barber:

Is that where it went! [Laughter]

Davies:

It was a very heavy thing, a teleprinter. I put it in the kitchen first of all, but it was so noisy I quickly got thrown out of there. So I got a piece of plywood and put it on the bed and stood it on the plywood. And I sat my son, who was I don't know how old, very small at that time, in front of it and showed him how to use it. It had a number of — this was using Scrapbook by the way, he quickly got used to Scrapbook — and he was going through trying out various programs that had been set up — I'm not sure, it could have been Scrapbook, but we were accessing these programs. There were conversational programs written by Chris Evans.

Now, Chris Evans was interested in applying computers to counseling, essentially. People who were ill, or perhaps mentally ill in particular, would use a computer to answer questions, not as a diagnostic tool but in order to go through a whole list of questions. Partly for the benefit of the medical people, but also he found — to his surprise — that actually talking about their problems often was enough, and that they were unwilling to talk to people but just using the computer [was acceptable to them]. This has been discovered since. He discovered this. And one of his things was on psycho-sexual problems. My smart boy logged into this, and he came up to me and he says, "Please, Daddy, what's going on here now?" And I looked at the program and it said, "Do you still enjoy intercourse?" [Laughs] I don't know how I explained it to him!

Abbate:

They would never allow that on the Internet today.

Davies:

Chris Evans was a remarkable guy. He wrote a book called The Magic Micro. It was a tremendous publishing success.

Barber:

He died — it was very sad really — he had kidney trouble.

Davies:

He was really a brilliant man.

Abbate:

So did you find that users were making, that they developed their own software for the network? If they had some application that they thought, "We really need this," would they write it themselves? If they saw a need?

Davies:

Well, they must have. Probably what they would do in many cases was to use the data processing service to write their programs for the mainframe and use the network to access it. You'd have probably a local mini or something — very often it was a DEC PDP-8—

Barber:

Remember there was the Edit machine, a PDP8, on the front of the mainframe that ran batch jobs, so they could submit these things through the network to Edit, which would run the batch job and give them results back. So that was kind of like an interface which we joined on. And I think that was really quite impressive.

Davies:

To the extent that they might need very complex processing for an analysis or something like that which wasn't immediately available to them, they would probably get someone in the math division.

Barber:

That's right, there was the math division, and then later it was the computer response section.

Davies:

Yes, there was a division that would help. I would imagine that probably all kinds of things happened, some enthusiasts around the place — certainly everybody who wanted to — would be able to learn BASIC, because that was how they normally used the on-line services that were available from outside. It's difficult to know at this stage, exactly. Certainly the network went on being a main tool for the NPL for — until when? — long after I left, I think it was.

Abbate:

Mid-1980s, I think.

Davies:

In the 1980s, late 1980s.

Barber:

Yes, mid-1980s, because I became director for communications for the Alvey Directorate[1] in 1983, and I actually put in a switch at NPL to allow Alvey people to get into the NPL facilities. So that I was still able to do that as late as that.

Use of Scrapbook beyond NPL

Davies:

Another interesting outcome of this was that the Scrapbook service, which among other things would allow people in different places to combine their efforts on generating documents — a fairly trivial thing now, but modern at the time — was bought by the European Commission in order to enable them to develop their annual report at Strasbourg and Brussels simultaneously, because they had to write an annual report for which people had to communicate between Brussels and Strasbourg. They actually bought the Scrapbook system simply for that purpose, to write their annual report.

Abbate:

When was that? Any idea?

Barber:

It was the only sale we ever had. Appleyard, wasn't it, he was the head of the DG3 Directorate.

Davies:

David Yates could tell us.[2]

Abbate:

Because I saw a reference, I was looking at some OSI stuff from 1977, and they refer to, "We could this for our own net Scrapbook." And I thought, well how did that happen?

Did you have any sense that having the network changed the way people worked in terms of, did they collaborate more? That's one of the things that happened with the ARPANET.

Barber:

I think it's hard to say. I think if there is an error that we made at NPL, and I don't know whether Donald would agree, was that we were very, very — I mean it was a marvelous time, everybody was enthusiastic, they would be working late every night because they were interested, so you had a really terrific bunch of people. But they didn't really care about, if you like, publicizing their work in particular or anything like that. And we used to have these open days, once a year, and I can remember one chap, he used to come around every year from industry, and he used to say, "I come round here because I get good ideas." And in hindsight, we never had a front office, we never had a publicity group and so on, and so really all of the things that we did — and there were some really quite remarkable things in speech recognition and things like that which went on, I mean I've got some of these pictures here — we might as well not have bothered for the, if you like, the credit that we got, you know what I mean. In terms of networking for example, ARPANET was a great big thing, and all we did was to build a local area switch. And it reflects, I suppose, attitudes in the society and all sorts of things. We enjoyed what we were doing and didn't think too much about spreading the gospel.

Abbate:

It seems like, in light of the Wilson "white heat of technology" and all that, and I see these references to sort of encouraging industry, was there an attempt to spin this off to industry? Or were industry people interested?

Barber:

I think this is a bit more related to my own background. I actually was seconded from NPL to Euratom, as it happens, as director of the European Informatics Network. And in fact this goes back to Donald, coming into my office, [3] I can remember it now, one Tuesday afternoon. He said to me, "There is a meeting in Brussels tomorrow and they want a technical advisor, something to do with networks. Can you go?" And I said, well, provided that the travel people can get me a ticket, I can do it. And indeed I got to this building exactly on ten o'clock, went in, and there was about half of a dozen people in a room, and I went in and I said, "Where is the UK delegation?" And they said, "Aren't you the UK delegation?" And that was it, and I had to play it by ear!

But after that, of course, came the beginnings of EIN [European Informatics Network]. They made me the chairman of a study group to put the thing together, and eventually I became Director of the project and so on. By the end of EIN,[4] in 1980, when I came back to NPL — well, I never left NPL, because I said if I am going to become director I want to have my office and my headquarters based there — but when I reverted to the UK, I got posted to the Department of Trade and Industry headquarters.

And after nine months up there I actually resigned from the civil service, and the reason was — and I guess this is probably much the same in most countries — that you had in your ministry for industry, or whatever you happened to call it, people who might have been technical at one time, but they get out of date and they start to pontificate about what should be done based on knowledge that is out of date. So they are actually making the wrong decisions. And as someone who was relatively new from technology, I could see it was wrong, and in the end, after nine months, I couldn't stand it anymore and I resigned. And then I went into industry and joined Logica. But after a couple of years, when the Alvey program was set up, I was seconded back into the civil service, as a representative man from industry! [Laughs] But I think this is a particular problem that we had over here. After Harold Wilson got in, and the "white heat of technology" and so on, shortly after that, in 1965 anyway, they amalgamated the Department of Scientific and Industrial Research (as it was, it should now be Department of Trade and Industry) and the Ministry of Defense. It was one great Ministry of Technology. And that was quite good, because for awhile I was on the promotion interview panels for certain grades, so I found out a lot about the Ministry of Defense and things like that.

Abbate:

Were they working on networking? I mean in defense.

Barber:

Well, not really. Sure, they were using networking, but it would only have been with modems on dial-up lines and that sort of thing.

Abbate:

I saw the reference to the code Grid 77. That's the only thing I have seen.

Barber:

Donald would probably know more about that, because by that time I was involved with the European work. But what I was trying to say was that because of that coming together of one big department, they wanted to focus amongst other things the computing industry. And that's why, when we tried to use the Plessey latest machine [for the NPL network], which just would have been ideal, we were told, "No, the Ministry policy is: we want to support ICL, you will not use that, you don't have to use that, etc." And of course when we wrote down the spec and looked around, the next best machine was a Honeywell 516. If we had gone for the Plessey XL-12, which was actually a much better machine, architecturally, than the 516, the whole story could have been quite different. But such is the way decisions are made in ministries.

Abbate:

I mean, were the computer manufactures at all interested? I know in the States, people like IBM and DEC were developing their own network systems and so they had a sort of strange relationship with anybody else, but did you get the sense that the manufacturers here were trying to develop their own [network systems]. Or were they interested in, "Oh, we could sell this [NPL network technology] as part of our system"?

Davies:

No. I think the manufacturers over here were much the same as IBM. There were really building big mainframes and making multi-access and time-shared systems. And the computer manufacturers themselves were not particularly interested in the communications side other than dial-up lines. To give you a good example, take the OSI seven layer model: the bottom three layers were in next to no time, the top layers were never ever really agreed — and in fact now, of course, Internet uses the old ARPANET protocols anyway, so who bothers with the [OSI] transport layer, and so on. And the reason why it was like that is because the PTTs were selling services, so they had to have a standard so they could sell services; but the computer manufacturers were all in competition anyway, so why should they agree upon any sort of standards which would make people able to use anybody else's machine as well? And that's life, isn't it?

I just checked up on Scrapbook. There were in fact nine Scrapbook sales, I was quite wrong: to Shell UK, National Water Council, two to the MOD (Ministry of Defense), British Telecom, Murray Clayton (whoever they were), a German publisher called Baumann, and to the Commission of the European Community. The European Commission one would have been sold to them in 1974 or 1975, and was used for them quite widely, what I mentioned about the annual report was part of it, but they used it for a great deal of things.

Collaboration and Security on Scrapbook

Davies:

Those are reminding me that Scrapbook, among other things, allowed you to make references from one document to another, so it was in fact a kind of hypertext. You'd have a highlighted word, which if you called it up would direct you to another document. So in fact — and it also had apparently a user response facility so you could turn a text into a kind of simple questionnaire, with "yes" or "no" answers or something. It didn't appear in the text, but it would ask you a question according to what you said.

Barber:

It was really the first screen-based system, wasn't it?

Davies:

Well, it was a hypertext system of sorts.

Abbate:

Did you have any sense that with these kind of tools, it encouraged people to collaborate more than they might have otherwise, between groups at NPL?

Davies:

Oh yes, certainly.

Barber:

Hmm.

Davies:

Or was it just because they had to collaborate and then they chose to do it this way?

[gap in recording]

Barber:

I really think — my problem anyway was that I was much more interested in building these things and making them work and so on. Probably, honestly, I really didn't care whether you could predict uses for them or not.

Abbate:

Well, it's the same with ARPANET. No one really knows. That's one of the reasons we are trying to find out. It's hard to necessarily get that information.

Davies:

I would have thought the way that Scrapbook worked actually promoted interaction because you had to take a definite decision to look up some of these documents. There wasn't very much in the way of browsing; in fact, on the whole browsing was somewhat discouraged, because people used to leave all kinds of useful and interesting information around. It wasn't the done thing.

Barber:

Well, you say that, but it happened a lot!

Davies:

It happened a lot, and in fact after EIN had been set up and there was a European connection to EIN, it was a great puzzle when one night some of us discovered that a document of his was being examined. It turned out it was someone across the Iron Curtain who was actually accessing it through the International Institute for Applied Systems Analysis at Vienna. It created a bit of a panic about the security at NPL.

That is because we set up deliberately a European, a Pan-European connection. I'm not sure if it was through Vienna or outside it, but anyway it connected to IIASA, and from there into some Eastern European countries.

Barber:

It would have been via EIN, I think. Maybe it went through Zurich.

Davies:

Well, anyway, it didn't have the feeling that whatever you had on Scrapbook was automatically open to everybody. If you wanted it to [be], you'd tell them about it. It wasn't. I'm not quite sure what the situation was when you wrote a document, whether it was open until you declared it closed.

Barber:

I think it was. Because I can remember —

Davies:

But you could also make it closed.

Barber:

But when I was negotiating EIN —

Davies:

Because if you wanted to make a document available to someone, sometimes the access rules make it only possible by sending them an e-mail. I remember that. The access rules would sometimes prevent you.

Abbate:

Did people use it for e-mail a lot?

Davies:

Oh, yes of course. Yes, a great deal.

Barber:

Oh yes, that's right.

Davies:

But then I had to clamp down on some kinds of use. It was being used for people's private clubs and things. People were using it as a means of storing information for private clubs.

Barber:

But there was a time when you could, because I can remember when I was negotiating my contract for the EIN, I remember Roger Scantlebury saying to me one day, "Do you know everybody is watching your contract being negotiated?" So I removed it entirely from the system!

Failure of Conferencing on Network

Davies:

So we had quite elaborate access control rules. You could use them to make it secret, or you could use them to make it public. The general feeling was that it wasn't intended to be public, unlike stuff that you would display. We did have conferences. They weren't a lot of success. Very early on, we had this idea of having set up conferences where people would join in. They weren't a great success.

Barber:

We showed it in the 1976 EIN International demonstration when all EIN sites showed the same set of services via the network. These had the conference system going in.

Davies:

I wonder why it wasn't a success.

Abbate:

So you mean [the conferencing took place] in real time, or like a bulletin board thing?

Davies:

Yes, it was a bulletin board. You wouldn't expect for us to communicate in real time. They could read the text and add something to it. To a certain amount it was used, but I wouldn't say that it was outstanding success. Perhaps it was the nature of the people at NPL that they didn't tend to have a lot of interests in common with others at the NPL.

Barber:

I suppose I'm skeptical. Even with Internet, for example, it's the sort of thing where if you really want to interact with somebody, you can't do better than to ring them up and talk to them, or something like that. This is the sort of problems you have with any, if you like —

Davies:

In the case of Internet, it gives you a much lower-cost method of communicating all over the world. But in the case of NPL, of course, since it was all in a small area, it wasn't necessary. The only point of the conference would be a way of bringing together the comments of a lot of people on a single subject. We tried it and set up subjects which were followed up for a while. I looked at some of the text but I don't think it ever lasted. I think it was done experimentally to see how it worked. I think in the end it didn't prove to be a success.

Barber:

Well after that, you see the European system, the Swedish system that was offered by the commission for a long time. And I tried to use that to support EIN, and again it wasn't that successful. I mean there were one or two enthusiasts, but you would get these endless comments on another previous thing, and you'd finish up with stacks and stacks of comments.

Davies:

You'd forget what you started with.

Barber:

Well, yes. Sometimes, and in any case sometimes you'd got something better to do than sit there and read it through a lot of comments. Many of which —

Davies:

Many of which aren't of interest. You can waste a lot of time reading rubbish. Many people I know keep an eye on particular user groups and so on just to see if anything is happening. But they don't read it carefully because there is too much of it.

Abbate:

So it's maybe less appropriate to a local area network like that?

Davies:

In a local area network it didn't seem to me that conferences were very successful in our particular case. But they were very early, and people, don't forget, most of — many people didn't understand what they were. It wasn't a concept that was current. So you can't say that it was unsuccessful, that there was some psychological reason for it, just because people hadn't bothered to find out what it was.

I am trying to remember the context or the subjects of some of those early conferences. I think Chris Evans did some of that actually, he tried setting up —

Barber:

I think he would have done so. Well, of course by that time I was too much involved with EIN to be involved in the divisional things.

Davies:

I'd taken part in negotiations in connection with a standard using computer conferencing in a sort of on-line fashion where we were all writing at the same time and then looking at each other's text and so on. It was a complete shambles. Nobody could get a picture of what everybody else was saying. There were about a half of a dozen of us, it was an international meeting, and we were supposed to be arriving at some views on some technical question. Because we were all writing at once, and some knew what others had written and some didn't, it led nowhere. It was an afternoon wasted actually. So that was my only experience with that. Mainly because it was uncontrolled; it needed a good chairman, I think, and there was no such function, we were all just writing our views on the last piece of text that we'd read. There was actually no use in doing it.

Barber:

Yes, you are quite right, it is the chairman in the end that makes a meeting either successful or not.

Davies:

In a case where there's a real potential for disagreement, the chairman can very often —

Barber:

Normally he lets everybody talk, and as long as everybody's said something, you go away and you —

Davies:

Well, but also you can discover what the real nature of the problem is and stop wandering about on the outskirts of the problem. But this particular case I had was a waste of time. I've had quite good telephone conferences with three or four people, fairly recently, and again they took a lot of time, but compared with traveling and having meetings and coming back they were actually a success. And that was — I think in those cases also one person had to act as a chairman. In one case, it was me because I was the only one who could hear all the others clearly! We tried to set it up with the Israel Telecommunications authority, and the technology was awful. We actually came back to Paris, and in Paris we did very well. It really worked extremely when we used it in Paris.

Early LANs and Ring Networks

Abbate:

I have one more question. What I'm thinking about, in terms of the historical significance of NPL, one obvious area is in packet switching, but I also see it in the context of development of local area networks, being the first one. And so I am curious, I had read somewhere that you were working on other local area network work and that you had maybe had a sort of prototype Ethernet-style broadcast system you are going to try?

Davies:

No. What you might have heard, there was an early Ethernet-type — well, it wasn't Ethernet, more of a ring network — yes, yes, I know what you're talking about now. When — Aloha was the first Ethernet, of course, that was a radio version — and at one point we conceived the idea of setting up a local radio Ethernet using a leaky cable as the transmitter.

Barber:

That's right. I got the cable put in, didn't I?

Davies:

Yeah. We were sat on eventually. We were not allowed to do it, because we couldn't get the agreement to use that particular frequency. We were told, what I was told, I think, was that —

Barber:

We were near London airport, it was going to interfere with the planes, or something ridiculous like that.

Davies:

No, what we wanted to use actually — in the end we wanted to use a television channel that was only used somewhere in Norfolk or something like that, and what we said was, "We'll only use a couple of watts or something, and we can't possibly interfere with them. And we are quite capable of handling, because of our error control, any noise that you might get from Norfolk. If the weather there happens to be bad there than the experiment will wait until another time, if the weather happens to cause the Norfolk transmission to reach us." So we were all prepared to do that. We had to get permission, from the Home Office I think it was at the time, to use this particular frequency. And we were turned down, we were told that under no circumstances do they license any new transmissions of that kind unless it was for mobile use, because they maintained that bandwidth over the air was so scarce that in the future they were only going to allow new types of systems for mobile use where radio was absolutely essential. They wouldn't allow any for static use, which was our case. What we wanted to do was to have terminals set up anywhere in the laboratory, communicating by an Ethernet -type, an Aloha-type transmission method, using the leaky cable as the sort of hub. It would have been nice, actually, it would I think anticipated some of the later work. I'm not sure if it was pre-Ethernet —

Abbate:

Well, Aloha was before Ethernet.

Davies:

I think it was between Aloha and Ethernet, actually, but I am not sure about that. It was [Les Pink] who was one of the people who —

Barber:

That's right, on the third floor, down from my office, down the corridor.

Davies:

We'd laid it all out, we had the thing ready, and we were unable to continue because of the authorities not allowing the frequency to be used. We should have gone ahead and done it, actually, no one would ever have known it. But we didn't.

[Tape is garbled here]

That must be a tough thing. There was another instance in which we may have helped to start Cambridge off in this direction. We were aware of ring networks that other people had developed, actually, they were in the literature but hardly — nobody seemed to have notice them. And Maurice Wilkes rang me out one day, and said, "Look, we are interested in the idea of local networks. What kinds of local networks have you heard of?" And I trotted out to him a number of different types of rings, and I sort of pointed out the interest there was in local networks and the way we had done it at NPL, and I think that may have helped him to start off, because it was before they started work at Cambridge. In fact the Cambridge ring was a different type from the ones that I tried to do. But I think it was an early influence.

Barber:

They came down — I remember them coming down to my office one afternoon, spending some time describing what they were proposing to do. And saying what did I think, and I said it sounded jolly good, and that was it.

Davies:

This was before, when he was talking to me he was trying to find out about what I knew of other types of local-area networks.

Abbate:

So they didn't already know about ring networks?

Davies:

They obviously didn't have the references that I gave them. I had picked these up from my general reading in networks, and some of those still are apparently obscure, not often quoted. I think it is an early — it's been so long I can't remember now.

Barber:

I think the paper's in here. [Looks through documents.] Newhall and Farmer, wasn't it? Here it is: Farmer and Newhall, "An experimental distributed switching system to handle bursty computer traffic."

Davies:

That of course would be in our first book, wouldn't it?

Barber:

It would as well, yes.

Davies:

Yes, that's why I still remember it. You only remember these things if your memory's jogged from time to time. Yes, it was Farmer and Newhall. Very rarely quoted as one of the pioneers. I don't know if they actually built one. I'm not sure.

Barber:

Yes [quoting from paper]: "An experimental switching system has been built, using conventional, etc., etc." That was the first paper. That was just the summary of it. And then there is [literature?] and so on. That was the contents.

Abbate:

[Looking at paper] So that's what this is?

Barber:

This is just the front page of that. Let's see if we can still find it. This is my actual copy handy there. Look, they weren't even [bound up?].

Davies:

That was Tom Lehrton?

Barber:

That was Tom Lehrton? That's right.

Davies:

And we found out about it.

Barber:

Yes, it probably was, actually.

Abbate:

Well, this has been extreme helpful.

[Davies and Barber start looking through old papers that they have collected.]

ICCC 1972

Davies:

This is my report on the ICCC 1972.

Abbate:

Where the ARPANET demo was? I'd be curious to see that.

Davies:

[Looking at another paper] Is this from Bell Labs? Oh, this is Spider, a local packet switching network called Spider. Fraser. You remember that.

Barber:

Spidernet? That was the Bell Telephone chap —

Abbate:

Yes, that's right.

Barber:

He was from the UK, actually. What was his name?

Davies:

Dr. Fraser.

Barber:

Was it Sandy Fraser? Anyway, Fraser it was. He was from Cambridge University, and went out there.

Davies:

[Looking through more papers] Crocker.

Barber:

Steve Crocker.

Davies:

[Reading from a report] "At the invitation of Steve Crocker of ARPA, Barber and I joined an ad hoc group on October 23, a public holiday, to discuss international standards."

Barber:

I don't remember that. [Laughs.]

Davies:

"During the luncheon I undertook to meet others and propose concrete action. Accordingly, we split into two subgroups. Group II was chaired temporarily by Vint Cerf."

Barber:

That was it. That was the beginning of INWG [the International Network Working Group]. Vint was the first chairman, and I was the next chairman, from 1976 to 1979, and then Alex McKenzie took over, I think.

Davies:

[Reading] "There were twenty technical papers, contained in mainly four of the twenty main sessions, were covered jointly by Barber and myself." So that's that. Interesting papers. Zacheropello  et al, IBM Zurich, on the Pierce loop. That's the one, the Pierce loop, not Farmer and Newhall. That's another one, that's another Bell Labs. You know [John] Pierce was the head of Bell Labs, and in his sort of spare time when he wasn't administrating or anything he invented the Pierce loop. I think it was only a theoretical thing.

Barber:

I can't remember. Well the name, you say Pierce loop, but I can't remember anything about it.

Davies:

Well, Manning of Waterloo University also gave a talk on "Pierce loops or Newhall loops", so I wonder if it is the same thing. [Looking at report] Remi Dupres. That's quite an interesting report, isn't it? [Laughs] I had a good time. I went to NBS [National Bureau of Standards], too.

Barber:

I must have written a report on it as well, but I don't know where it has gone. Well, actually I've kept [a lot of papers] — yeah, my wife goes mad some times in my stuff, with all the stuff I got.

Davies:

Ah, I've got strict orders: no one is to be allowed in my study, it's such a mess!

Barber:

Well, do you think we have covered all that you want?

Davies:

I'm going to make some tea. Would you like a cup of tea?

Abbate:

Sure. Well, I think that we certainly have covered a lot. I will probably have some more questions at some point.

Barber:

What's the actual objective?

Abbate:

Oh yes, I guess you missed the beginning [of the interview].

[Stopped recording, since this was going to be the end of the interview. Over tea Davies and Barber began talking again, and the interview resumed informally.]

Interview, Part 2

International "X" Standards

[The following seems to be about the PTT's development of the "X" standards for data networking.]

Davies:

— in order to establish what is needed. And the two people who sponsored this, who actually put up the motion and seconded it, were the Russian and the Canadian delegates. I remember that, which was rather strange that the Russians should come in. And so they set up GND, Group … no, it was Reseaux — special data networks — NRD, Nouveau Reseaux [D'ordinateurs? see also GMNRD below], NRD, which was a special committee for the next session. And during the whole of that four years all it managed to do — these were nice meetings, I enjoyed them in Geneva — but all we managed to do was to develop our first standard, which was X-1, which is requirements for packet switching networks. Just to set down about twenty or so requirements for packet switching networks.

Barber:

Two things: I know who you mean in the P.O. [British Post Office]. I can't think of his name, I know who you mean. And the other thing is that thinking about it — it was probably a good thing that Fred [Wharton?] gave that presentation. Because I can imagine the impact of that on the PTT people, there is IBM trying to say, "We will do it all," and then Donald comes in after the break, saying, "Look here, chaps," and we wound up doing it! [i.e., the PTTs wound up making the standards.]

Davies:

Yes, well, it's for us to develop that, you don't want to put it all in the hands of [the computer manufacturers]. I don't think this guy in question [from the P.O.] — I don't think that he was at this meeting.

Barber:

No, he wasn't. His name may come to me, but it was a bit before my time in the Post Office, but I know who you mean. The reason why I'd like a copy of that, actually, is 'cause I'm still in the retired staff association at the Post Office.

Davies:

I might have a copy. I've got multiple copies of many things.

Barber:

Thank you very much.

Davies:

Now, in this next session they did X-1, and in the next one they did several, X-2 and X-3. These were setting up more details and then in the next session they also started to set up a committee to develop an interface, and so in the following one after that X-25 came out. And it was largely — the leading people in CCITT were Larry Roberts, who had joined by that time, and the French, who were already engaged in developing their network. They were the driving thing, because they were beginning to develop a network and took a lot of interest in Ê.Ê. [Typo?]

Barber:

Yes, because you had Louis Pouzin and Remi Dupres. Yes, Remi Dupres, because he was actually in the [French] PTT.

Davies:

The multi-layer version, the multi-layer structure of X-25, I think was largely due to Larry Roberts. It was very curious to see Larry Roberts in this new environment, because CCITT has a very formal way of working. For example, whenever you get up to speak you always thank the previous speaker, regardless of whether he was talking absolute rot. You must thank previous speakers. You don't just jump up and talk, you have to wait to be recognized by the chair, and very very formal rules. Larry Roberts didn't sort of see all this at first, and at first no one would talk to him because he wasn't one of the boys, and it took him a long time to get into the [knack of it], but eventually he was very effective. [Laughs] So that started off the PTTs in this area.

Influence on ARPA

Davies:

But that's really by the way, because we were really talking about the way in which we might have influenced ARPA. Now, we fairly early got onto the ARPA network, and throughout the early development at ARPA we had continual interaction with them, this was even I think before BBN was brought in. We already had started to discuss things with them.

Barber:

Well, yes, because Roger [Scantlebury] was at that first —

Abbate:

That was in 1968?

Barber:

1969 this was.

Davies:

There must have been one in 1968 as well.

Barber:

Well, I can't remember that.

Abbate:

Was there in Scandinavia somewhere?

Barber:

There were meetings in 1974 in Scandinavia, that was the CC. There were two, one after the other. IFIP was one week and the ICCC was the other week.

Davies:

Well, in 1968 of course was the Edinburgh conference.

Barber:

That's right.

Davies:

That was — our big publicity came at the Edinburgh conference. [Looking at papers] Good lord, a letter from John von Neumann stuck in here!

Abbate:

Did anyone from ARPA come to NPL, I mean to look at it?

Barber:

Well yes, we had that meeting in — was that in 1966? Couldn't have been 1966, could it.

Davies:

Well, 1965 was the one where we started it.

Barber:

Well, we had that meeting in — in Building 93, and Larry was there.

Davies:

Well that was in 1965, that was before ARPANET was even started. It hadn't begun.

Abbate:

I mean, did they come to see your network?

Davies:

Yes. We had individual visitors, I know. I can't think if we had a kind of delegation.

Barber:

No, I don't think that we had an actual delegation.

Abbate:

But even individual people came to see it?

Davies:

Oh yes, individual people, certainly. They knew what we were doing, we would exchange papers regularly.

Abbate:

Do you know who came?

Davies:

Bob Kahn was one of our main contacts. He was the — Then Alex McKenzie.

Abbate:

Oh, Alex McKenzie, did he come? I couldn't imagine that.

Davies:

I can't remember, I can definitely remember that Bob Kahn — There was — Another thing Roger [Scantlebury] could tell us, actually, because Roger had a party one evening, after a conference I think, in his house, in which they all sat around inventing an early version of —

Barber:

TCP.

Davies:

TCP.

Barber:

I think that was an IFIP meeting, wasn't it, because it was Alex Collin, wasn't it, who was a Canadian? He chaired that one. Vint Cerf was there. On that committee there was Vint, there was Roger, there was Michargien, and I can't think of the fourth. The four of them actually laid the first beginnings of TCP/IP. It was IFIP, 1966 I think.

Davies:

All I remember is that there was a party going on, and instead of joining in the party, a group of four of them had disappeared into the back room.

Barber:

That was the group of four.

Barber:

I would have gotten my diaries and I could have looked into that. It probably was 1966.

Davies:

Oh no, later than that. It must have been later, because ARPANET must have been in full working order before they started to work on [internet protocols].

Barber:

Oh yes, of course, I'm not thinking.

Abbate:

Probably 1970.

Davies:

1970-something.

Abbate:

But that was here?

Davies:

It was in Roger's house, actually.

Barber:

Oh yes. Because I know that it was held in — I was already director for EIN by that time, because it was held in that building in one of my — I really was — Yes, that's where it was held. Yes, so that would have been 1972 or 1973.

Davies:

I know that when the first bids came in for the construction of ARPA network, we almost immediately had a copy of —

Barber:

That was in 1969.

Davies:

Whatever it was.

Barber:

Yes, it was. Because I can remember coming into your office, and you'd got this copy of this report —

Davies:

And it turned out that they were using the same computers that we were [i.e., both networks used the Honeywell 516 for their switching nodes].

Barber:

That's right. But it wouldn't have done if it hadn't been for our rotten ministry, who forbade the use of — [Laughs] Remember how they were trying to rationalize computer types, and we wanted — There was a very nice experimental machine, very fast machine in the UK that had been developed, and we were going to use that. The Plessey XL-12. It would have been ideal, and we were going to use that, and there was an edict from the ministry to say that, "We are trying to reduce the number of computer manufacturers" — or words to that effect — "so we don't want you to use that." And of course that was it. And in the end we cast around and the next best thing was a [Honeywell] 516, so we had the 516, which is of course exactly the same [as in the ARPANET].

Davies:

I remember it well because I was... [Laughter].

Abbate:

Can I ask another question?

Davies:

This is a report from Roger Scantlebury on this 1967 visit [to the Gatlinburg conference]. There are three papers, he says: "The second paper by Dennis, the NPL paper was well received, and the ARPA network is being implemented using existing telegraphic techniques simply because the type of network we describe does not exist. It appears that the ideas in the NPL paper at this moment are more advanced than any proposed in the USA," he said in his report.

Abbate:

I might have a copy of that. Let me see if I can tell —

Davies:

I'll make you one just in case.

Abbate:

No, I don't think that I have this.

Davies:

Let's see if there's anything here. Oh, Llewellyn was the man.

Barber:

Arthur Llewellyn.

Davies:

Arthur Llewellyn was the man from —

Barber:

The DTI.

Davies:

No, it wasn't at that time. He had just come out of MOD [the Ministry of Defence].

Barber:

Oh, had he? I hadn't realized that.

Abbate:

And who was he?

Barber:

Arthur Llewellyn his name was, and as Donald said, he was originally from MOD and moved into — Oh, I know why it was, it's because the Labour government combined MOD and DSIR into one great ministry called the Ministry of Technology. That's what happened, so that's local UK politics, really. But it had an interesting effect, doing that, actually, because you brought together a lot of people who would never have talked to one another.

Davies:

Here it says: "At the end of 1967, I attended the final meeting of the special study group A of CCITT. It was formulating the recommendations that were going to the plenary in 1968. In that meeting Fred Wharton of IBM gave a long lecture on the future of telecommunications from the industry point of view. The chairman of the study group, Mr. Rhodes  of the UK Post Office, knowing of my proposals for a new data network, approached me during the interval to ask if I would like to give a short lecture on the principles of packet switching. I was able to introduce packet switching to CCITT. A resolution was passed for the establishment of a joint study group on new data networks in the 1968 to 1972 session of the CCITT. This is called GMNRD (Group Mixed Nouveau Reseaux [Dinari?]). GMNRD produced the first recommendations for new packet networks in the form of lists of facilities and introduced it as a subject for study, for the detailed study had to wait until the period 1972 to 1976." But the first time that we really were able to describe our work in full was the IFIP Congress in Edinburgh in 1968, where I think we gave four papers.

Davies:

Four papers, there were.

Abbate:

Yeah, they had those [at the archive].

Idea of General Purpose Network

Barber:

That was where we were really sort of able to describe things, both our concept of a general purpose network and what we were doing in the local network. In the discussion — I don't know if it is in the papers, there was not very much discussion after one of the papers I gave, except an IBM man — not IBM, an ITT man or a Bell Labs man, I think it was — who got up and said, "Look, what you are saying is very nice, it's very idealistic; you won't have a chance at making it work. We know that any kind of switching system, the software is incredibly complex, and you simply cannot make it work at the kind of speeds that you are talking about." [Laughter]

Well, that was Number 1 ESS [Bell Lab's first electronic switching system], wasn't it? Bell Labs had a hell of a job to make it work. [Laughs] That was their problem.

Davies:

The point is, if you tried to make a message switching system work, because you were trying to imitate everything that has been going on from the days of torn tape until present — all with people doing it — you get yourself into an awful mess, and it hardly ever works, certainly not at the speed that you expect. What we concluded, what we contended, was that if the switching principle is extremely simple, the chances are that you can make it operate very fast.

But we never really believed that in the long run a fully blown network would consist of general purpose computers doing this. One would imagine, like the telephone network, that it may have general purpose computers as part of it, but there would be a lot of specialist hardware. And so really what we were looking forward to was something more like ATM, where the detailed switching is done much better on hardware controlled by computers. We regarded the idea of using a general purpose minicomputer as a switch as an intermediate stage, but nobody at that stage or since has ever taken up this idea and developed special equipment. There was one small effort to that in Canada, at the University of Waterloo. The chap there actually developed or designed — I'm not sure he ever built anything for it — specialized packet switching equipment, but it never got anywhere.

So in effect what happened was that the ARPA network brought packet switching up to quite a high standard of technology based on the fifty kilobit line and the thousand bit packet, but there it froze, and that was — Again, the X-25 [standard] was marvelous in the sense of getting it [packet switching] widely spread in use, but it also froze the technology, because everybody said, "That's the standard, you must stick to that." And consequently it never moved forward, there was a big hold up until ATM took over. And ATM represents the next big leap forward. Of course, there are actually many variants of packet switching which make it faster. For example, giving up the idea of error control at the inter-link level, at the inter-node level, can help a lot, and that's happened in cell switching, as you know. And this is being discussed in packet switching; in fact it has happened. There is also what's called "cut through," where some packets don't actually get stored at all, they go straight through. All kinds of tricks are being developed to speed it up, but none of them went very far. I think cell switching and ATM is the essential thing.

Switching for Telephone Communication

Davies:

We also, by the way, right from the start — well, from the very early papers, and I can't actually find the one now — I felt that a packet switching network, when it existed, will be the ideal thing for carrying the message traffic of the telephone network. Because the telephone network, as you probably know, in setting up calls, has to exchange lots of short messages between exchanges. What do they call that stuff, the switching traffic?

Abbate:

Signaling?

Davies:

Signaling, yes. A number 6 signaling system was being developed at this stage, and my view was that you shouldn't develop your own standards here, you should develop a general purpose packet switching network and then provide an interface into the telephone system by which you could carry the messages, so it seemed to me that packet switching was an ideal carrier for switching traffic. But beyond that I did consider the possibility of telephone traffic, and with the technology at the time it looked a little bit far out, but it was definitely a possibility for the future. And in order to show that a little bit, we set up —

Barber:

— a demonstration on our local network.

Davies:

So we went along and found delta signal modulation was about the simplest thing.

Barber:

I remember buffering it and stuffing it through our network, and it worked all right.

Davies:

We had a tape with "Mary Had A Little Lamb" on it, and the funny thing was, at the end we were employing some interface, which controlled the flow of packets — the flow of characters — out, so when you switched it off, it didn't throw the characters away. It stopped the line. And so the buffers would fill up. And what we didn't know was that there was so much buffering in the network that you could switch off for several seconds your terminal, and when you switched on it started immediately from where it left off. And we thought, "God, we never thought of that!" We thought of it like a telephone system.

Barber:

The thing that spun out of that, sometimes afterwards — you know when divers are down and they breathe in helium; well, the technique that is used to make the voice come down to the ordinary band is my technique of buffering, and then feeding out much faster than you feed it in.

Davies:

Yes, I guess so.

Barber:

It was! We never got any credit for that. [Laughs]

Davies:

What we did, actually, we had to have a clock pulse. You see, the network is asynchronous, so we had to have a clock pulse that was known all the way around the laboratory where you could listen to these nursery rhymes. So there is a very powerful long wave transmitter which you can pick up — it's so powerful that you can listen to it on a piece of wire like that with headphones, you don't need an amplifier. So we used this as our clock pulse, in order to be able to regenerate the speech at the right speed. But of course, if your buffer was full you wanted to get rid of it, so you had to speed it up a little, and no one would notice if we speeded the speech up a little bit, to get rid of the buffer. That was quite fun. It just stopped there, but we just wanted to show that voice traffic could be carried on a packet network.

Abbate:

One thing I'm wondering about: it sounds like the British Post Office, or at least maybe just the engineering department, were very accepting of this, and they said, "Well, sure, this is message switching, we understand." Whereas it sounds like AT&T — and Baran and Roberts say the same thing, that AT&T was just very skeptical — [saying], "It would never work." Do you have any idea why that would be the case?

Barber:

To be fair, there was only part of the Post Office that were like that. I mean, internally, I think the people that we convinced actually had quite a lot of problems with the die-hard speech people.

Abbate:

And they thought it wouldn't work, the others?

Davies:

I don't think it ever had really strong support from the top. There was a part of the Post Office which liked the idea, and therefore EPSS never had the full support that it could have.

Barber:

In fact, they were the data people when you think about it. People like, who was before — Anyway, what I am saying is that people who had to deal with data and getting it through the telephone network were the people who were, if you like, sympathetic to this new approach.

Abbate:

So the AT&T people weren't those people?

Barber:

I don't know, I wasn't that close to AT&T, but I remember going to — I can't remember when — it was a conference where there were a lot of AT&T people there. And there were one or two of them who I thought were very sympathetic about packet switching, so I guess it was probably the same. The majority would have been against it, but there would have been people there who were, if you like, more open-minded, different background perhaps.

I'll tell you something that I have always found amusing in hindsight. It was 1969 that I went to the States to this, but I visited various other places, and I went down to AT&T down on the bottom of Broadway, 195 Broadway or something like that. There was a chap there, I don't know, [Serge Bowman?] was his name, that's who it is. And they were working on what became the standard RS232 interface. At that time I'd got this interface of my own here, and all I'd got was basically four controller wires: a pair for handshaking the bytes and another pair for saying whether the ends were actually to interact, and that's all I'd got. And I remember going there, and if you look at the RS232, it's got twenty-five wires! And not only that, mine was [ten?] bytes; you'd only got two serial data — [this part of tape is garbled] — so you got two wires come in later, and twenty three other wires. The AT&T people apparently say, "We don't know what these computer guys actually want. So we thought if we gave them plenty of connections …" [Laughter] Really, when you think about it, I don't know anybody who has ever put two bits of kit together with RS232, the American standard, without not knowing which way to cross the [wires?]. So there you are: that's another anecdote.

Scantlebury's Influence on Baran

Davies:

Here's an interesting point that Katie Hafner makes, which I didn't know. It says that at the Gatlinburg meeting Roberts heard for the first time about Paul Baran's work from Scantlebury.

Abbate:

Yes, that's what he [Roberts] says.

Davies:

I didn't realize that. [Reading from Hafner:] "Whereas the Ann Arbor meeting months earlier" — that is talking about this — "had been the intellectual equivalent of a barroom brawl, this was high tea at Fortnum and Mason's." Katie Hafner is writing a book about the network. "People were politely coming around to the idea of a network. So Robert's presentation was generally well received, even enthusiastically by some, and afterwards a colleague of Donald Davies's from the National Physical Laboratory named Roger Scantlebury approached Roberts and told of the work going on in England on packets. It was the first Roberts that had heard of it. Scantlebury also raised the issue of line speed, etc. etc., ten times faster than what Roberts was proposing and also told him about being the work of Baran. When Roberts returned to Washington he found the RAND reports, which had been in his office unread for months."

Abbate:

That's what Baran says, that he [Roberts] had them before, but who knows if he read them.

Davies:

[Reading from Hafner] "Roberts was designing experimental networks not with survivable communications as his main or even secondary concern, nuclear was not on his agenda. But Baran's insight into data communication intrigued him nonetheless." This guy that I used to work with — he became quite senior in Canadian communications — he told me that he was the supervisor of Baran at that time, and that he gave Baran the idea. Quite a well-known name, in the telephone business more than data, I think. He was Baran's supervisor.

Barber:

Was he from Bell Canada?

Davies:

Yes, I think that's where I met him. I certainly met him in Canada. Eckerton, I think was the name.

Barber:

It wasn't Eckert —

Davies:

I shan't remember his name. And that surprised me, and I was a little bit skeptical when he said to me, actually, but that's what he says. I don't want to give you anything of Katie's because she might not like it, but that's the only bit that was news to me.

ICCC 1972 Demonstration of ARPA

Barber:

What I didn't find out was what your actual interest was, was it in the very beginning of the work?

Abbate:

Actually it's in quite bit of it. I am interested in the beginning. One thing that I did want to ask about was when you had actually built it and people were using it, because one of the sort of unexplored questions seems to be: What these things are like to use? Historians tend to look at when you build the computer, and then they walk away and they don't really ask about that. And you seemed — in particular when I read early the stuff — it seems like you are concerned with the ordinary users, as opposed to the Americans who tend to be concerned with the computer expert is going to be doing.

Davies:

Well I think that you are right. The emphasis in the very early ARPA network is on getting the technology right, correctly of course; they were building an elaborate device, and they had to get it right. It was only the — what was that meeting in Washington, the one which was the very first of the ICCC?

Barber:

1972.

Abbate:

ICCC.

Barber:

ICCC in 1972, when ARPA put the network on display. I was there.

Davies:

How did we both get there?

Barber:

I was there because I was funded because of EIN.

Davies:

That's right. Unusual for us both to be at a conference at the same time. Well, that was the first time that I saw ARPA network as a user network, because everybody who was there was actually trying to use it, and mostly having great trouble with it trying.

Abbate:

Really?

Davies:

Well at that time everything was very experimental.

Barber:

Well, they pulled out the stops to try to get something there at all!

Davies:

It was very impressive. I mean that was a really watershed. Before that time, before 1972, if you talked about packet switching you were always on the defensive, saying, "Well, it's very good in spite of what you say," and so on; everybody was against it. After that, it was quite the other way around. You had to hold people back and say, "Look, it's not quite as marvelous as you think it is." It was very important as a piece of demonstration that the whole thing could work, and it was really, I think, what impressed me was the amount of intellectual effort that had now been attached to the question in US universities. After that point there were very few people —

Barber:

They'd already got the first four nodes, more than the first four nodes in, hadn't they? And they'd already got people switched on in the universities, because they'd got nodes to the network, and I think that is what did it.

Central File Storage

Davies:

Well, we were concerned right from the start, in our first discussions on this, when we discussed how we would go about this, and decided to do the simulation work — which we haven't talked about, but led to quite interesting papers — and also the local network, we said, well it's not any good just putting down the local network, that has got to be something that you can use there. People are not going to talk to each other just for the fun of it. So we decided that one of the most important things was to have some memory attached to the network, because at that time if you had a minicomputer you got a minimum, a very small amount of tape memory, very slow and so on. If you wanted disks they were very, very expensive, they cost more than the computer, but they gave you more capacity than you wanted. So what we wanted to do was to get a fairly expensive collection of disks to attach to a service to make a file server and attach it to the network. We didn't call it a file server, we called it a central file store.

Barber:

Central file store, that's right.

Davies:

So we had — Alongside the Honeywell machines that were doing the switching, there was another set of the same machines with as much file storage, as much disk storage as we could afford to build in. It took a little while to develop, but it actually proved to be very valuable resource.

Scrapbook and Edit

Davies:

But quite out of the blue, and without us promoting it, another service turned up which was called "Scrapbook," and that was a real user enterprise.

Abbate:

Who developed that?

Barber:

Well a group in another part of the division did it on their own really. It's really a combined — because in those days even word processing wasn't something that you just had, it was something that you had to work hard for — so it combined word processing, but more than that, it also provided an electronic mail service for its users and a means by which two or three people could collaborate on the production of documents. In fact it was an extremely effective service which worked — I am not sure if it is still working, is it?

Barber:

Scrapbook? No it was shut down.

Davies:

No, because most of its components are available in other ways. But it was a very valuable service. The other main service was the ability to collect experimental data, store them somewhere, and then bring out at the appropriate time and shove then into the mainframe. Because the mainframe wasn't providing on-line services, it was providing batch services.

Barber:

This is the Edit machine, in front.

The basic function of the laboratory is measurement science, as you know; a bit like the National Bureau of Standards, but somewhat smaller (or NIST, as you say). And therefore there were quite a lot of people doing experiments which produced enormous quantities of data. And we had to collect that and process it, and up to that time it had been done in specialized machines and collections of hardware, and was generally a bit of a messy and very expensive process. As a result of this [the network] people were able to do it much more systematically.

Photos of NPL Divisions

Barber:

[Looking at pictures] There is an actual picture of — I have got several pictures here. This is one of the terminals, see there's the four buttons that we had to log on, and that was in fact one of the girls that was in the Scrapbook group.

Davies:

Four buttons was all you needed to log on, to get through the switch.

Abbate:

Yes, I have seen the descriptions of it. I know what they look like. It says "Open day." Was this some sort of demonstration?

Barber:

Each year you had an Open Day there, not exactly entirely open to the public but to any technical people, and we had the network running quite well by that time.

Abbate:

So was 1971 the first one that you demonstrated the network?

Barber:

I can't remember. [Looking at another picture] That was the switching center by the way.

Abbate:

Is there any way that I can get copies of any of these?

Barber:

I haven't the negatives, I've only got these. The only thing would be to get the — Some of these are NPL ones, you might still be able to get those.

[Looking at another picture] That was the Scrapbook machine. That was based on a Modula-1 computer. Then there are a whole lot of things like this. These were the other parts of the divisions: speech recognition, pattern recognition, character recognition. That was the meeting that [?]. The amusing thing was that Brian [Oakley?] was there and I was there later because Brian was the head of the Alvey directorate and I was one of the directors in the Alvey directorate. It just shows how — And that was when EIN started and that's when I almost got — And that was the standard interface, the work we did before, which laid the foundation for the ease of getting into the network.. That was the network. Wish I'd got a date on that. That was either 1967 —

Abbate:

So if I had this number [on the picture], I could ask them if they still have it?

Barber:

I can do those copies now to give you an idea of what the pictures look like.

Abbate:

That will be useful, because I will be back in London in August.

Barber:

OK, we can certainly do that.

Abbate:

Is it possible to go see NPL?

Davies:

Yeah, it is probably more difficult now because they've become privatized to some extent.

Barber:

There was another thing that we had, actually, which was an infrared data link between the roof of our building to the roof of the computer building so that we could send data to the central computer without going through the cable.

Davies:

We also had a microwave link to the Ship Division.

Abbate:

That's the one at Feltham? What was that for?

Davies:

Well the Ship Division is one of the divisions at the laboratory that happens to have an experimental tank, and their only connection was to [the] network was through this microwave link.

X-Y Digital to Analog Converters

Barber:

Now that reminds me of something that is quite interesting. That was when I was involved with the process control work and did all the instrumentation which had been the foundation for that. And that has an X-Y display on a cathode ray tube. Now I went to MIT in 1963. Now, at NPL for a long time on the Deuce computer there had been a display of the contents of the delay lines, and by some very clever programming, really, people had made the delay line display screen have a ballet dancer which danced around, and it was all done by altering contents of the delay lines, which is a tour de force. [Laughter] But, I could never understand this — nobody had thought of doing what this was. And I went to MIT, and the first PDP-1 had just arrived and I saw that, and I went to see project MAC and met Robert Fano and all the people there, and they had on the PDP-1 an X-Y display, with X and Y digital to analog converters that displayed. Well, why did we never think of that? So as soon as we got back we built one. An example, I always think, of how you can miss the obvious way of doing it by being too clever.

Abbate:

Well, let that be a lesson.

Barber:

Do you have a copy of that paper from Gatlinburg?

Abbate:

Yes, I have that.

Davies:

That seems harder to come by. And you have gotten our papers from anybody?

Microwave Connection to Feltham

Abbate:

Yes, we have those in the archives. They have a fair selection of the published ones. The connection to Feltham, that was microwave?

Davies:

Yes, a microwave connection. It was quite funny because we wanted to put a tower on the top of the building, an aerial facing — no, it was a cable. I wanted to use microwaves, the equipment existed, and we got up onto the roof of our building and had a look, and we could see, across the tops of the trees, quite clearly the top of their building, which was a few miles away, and we said, "We'll put up some aerials here and fix up a microwave link." So I asked the director of the laboratory, and he said, "No, not at all. Anything which affects the exterior appearance of that building has to be approved by so-and-so, and people are really fussy about the park." And we were nowhere near the park, actually, and since then people have put up all kinds of horrible excrescences, but we were turned down. So in fact it was a cable link, it was a sixty-four kilobit cable.

Abbate:

And you also were linked to Harwell, the Atomic Energy —

Barber:

It was really when EIN started. That's when we got linked to Harwell.

Abbate:

I saw a reference, but I didn't know when that was from.

Barber:

It would be early 1970s.

Datagram vs. Virtual Circuit

Davies:

You know the controversy between the datagram and the [virtual circuit].

Abbate:

Yes.

Davies:

You really know a lot about this already! We were always saying that datagrams are fine because you can carry out enough interaction between the ends to control congestion, and provided everybody behaves everything's all right. And the communications people used to say, now that's terrible, we can't allow that, someone will make a mistake with a packet or something and get themselves into a horrible mess. And we never could resolve this problem. Well, EIN was set up as a datagram network —

Barber:

Well, no, what happened was, it was more complicated than that. I, of course, chaired the committee that tried to plan the network, and after a while I became aware that we were actually getting into a deadlocked situation. Because there were certain people from France who believed that the datagrams are the solution and that virtual circuits are absolutely ridiculous, but there were some other people, there were the Italians, who for some reason or other thought that virtual circuits were the way. And every time you proposed a facility like a single connection — no, duplicate connections to a host in order to provide reliability — you would get people who would want to set up a circuit and — Actually I should lose the argument. Anyway, to cut down on the detail, the fact is that every time someone proposed such a facility, someone else proposed a facility that wouldn't work if it was using datagrams, or vice versa. And in the end I thought, I can't get anywhere. So what I proposed — which broke the deadlock in the specification of the switch — we asked for all the facilities, so they had to be provided, but that the user could elect to use certain facilities, and that broke the deadlock. So though we provided virtual circuits, we never used them.

Davies:

Oh, you never used them.

Barber:

No, as far as I know we never used them.

Davies:

But we had a big sort of a demonstration at the NPL, in fact all the EIN centers were going to have a demonstration, for which they had rows of VIPs sitting in front of the terminals who were going to want the thing switched on and operate. We had the network control center at NPL. Then just before the thing started everything crashed, nothing was happening at all, and we couldn't understand this. I rushed down to the control center. The first theory was the packets that were flooding the network were coming from somewhere, and we found that was wrong, but eventually we tracked them down to Zurich.

Barber:

That's right, the Swiss.

Davies:

Zurich was pumping packets into the network as fast as it could go without taking any notice at all. It turned out that they had done this deliberately to try and sabotage the network because they believed that the datagram principle was wrong. They were sure that it was wrong. So all we did was cut them off, a flip of a switch and they disappeared from the network, and we continued the demonstration! [Laughter]

Abbate:

When was this?

Barber:

It was just as well we had the control center! That was another Modula 1. It was in 1976. I've still got the glossy brochures because I didn't know, [?] in our time scale. John somebody, the lad who did that, a really good —

Davies:

It worked very well. Another thing that we did early on was to provide a connection out to the IEE. The IEE of course was in the center of London, and we were going to give some talks about the NPL network, and in fact have demonstrations. We had to move out lots of equipment there and provide a connection to NPL through a PCM connection, 64 kilobits, which would be temporarily laid on for the purpose so that we could demonstrate the network at the IEE. I was going to give my talk with my lecture notes coming up on a screen using Scrapbook. Push a button, up comes the next page. That was fine. We were told the BPO provided this service free of charge on the basis that it was a service to be provided using a spare cable, but that cable was going to have to be available in case there was a breakdown somewhere of the network, so that they could rapidly switch it in, but we prayed that it wouldn't happen. Just at the point where I began to lecture, the connection was completely broken! Fortunately, the demonstrations had all taken place. The lecture was after the demonstrations, but my notes disappeared. Anyway, I had a spare copy. It just happened that they needed the cable, so they took it away. But it was a good demonstration that you could extend the NPL network to the IEE if you wanted to.

Festival Hall Meeting & Real Time Club

Barber:

We had something that was at the South Bank somewhere —

Davies:

Would that be the famous meeting at the Festival Hall? A pressure group was set up —

Abbate:

That was the Real Time Club?

Davies:

The Real Time Club, yes. They had a sort of public session at the Festival Hall, with a very powerful talk by Stanley Gill. Which was sort of saying what would happen if we didn't — You've probably got that, Stanley Gill's speech at the Festival Hall.

Abbate:

I don't actually have the speech. I know the story.

Davies:

The Real Time Club ran on and I went to their meetings. One of the people who ran it, one of them ran a company that was providing information services and it became bankrupt, ran into financial trouble. I am not sure that it exists now. I don't think it does.

Abbate:

At the Festival Hall, was there a demonstration of the Mark I or the Mark II, whatever it was then?

Davies:

Did we demonstrate at the Festival Hall?

Barber:

Yes we did. We had some display terminals up there. I am trying to think of the very senior Post Office chap who came down there. Yes, we definitely had our things out there.

Abbate:

I mean did it have — Did people come and say, "Wow, this packet switching really works?"

Barber:

I think the problem is: what's the difference between having one of these things at the end of a packet network and having it connected by a line into your mainframe just next door? What we were showing was actually pretty good, using a network and so on, [but] as far as the end users are concerned, it didn't look any different from having it patched right into the mainframe in the next room. So only if you knew what you were talking about, as it were, did it have the right impact.

Abbate:

So there weren't any Post Office people who came and saw it or anything?

Barber:

Well, I don't remember. It's funny how you just jump around, but in 1976 at the ICCC meeting in Toronto, I had a display of the UK Teletext with the cricket results coming up, and it was done from — Nowadays of course Teletext is done by a chip inside the television, but we had a six-foot rack of equipment which [Alan Davis?] had built just to do it all. And we piped that — the RCP network, the French PTT, had got a link to the Toronto conference center, and we piped this thing from NPL through the EIN network into France, from France into the French RCP network, over this link into Toronto, and eventually down — I'm not even sure that it didn't go through the Canadian packet switch — but eventually it came down to this display, and those who were in the know were coming by to find out what the latest cricket results were. [Laughter] So that was, by 1976 we were doing Teletext.

Sabotage by the French

Davies:

So the French were actually helping you demonstrate Teletext? That's unusual because there was a famous conference at the UNESCO building in Paris. And just down the road down from UNESCO was the head of the PTT there, in the neighboring street. So they had arranged a small exhibition there in which people could come along and show their equipment. And the Post Office had brought along — was it Teletext, was there another name for it? Oh yes, Prestel. Which at the time was quite something to show people how you could get into a network and get information from it. It was trivial by today's standards, but it was quite something then. The French had nothing like that, and they were very envious of this and the British Post Office were showing all this stuff. The French were looking at it, and suddenly the line died. I was there with one of the Post Office people, and they immediately found that the plug had been pulled on them. People around the place had spare connections, so they said, well, give us one of yours and hook it in and we will get going again. So they did, and after a short while that line died! So they didn't realize it was really sabotage going on.

Eventually they got to the point where they couldn't get anything at all, no spare lines. And they were saying they were desperate, they didn't know what to do. So I said, let's go and find — this must be happening at the local exchange, the local PBX, so let's go and find it. So we marched down the corridors of this very big building, asking where the — he didn't know any French, and I knew enough French to be able to ask — and eventually we found the exchange and they said, yes, it was their instructions to pull these lines! And so he said, "Who gave you them?" And they told us the name of the guy. And we stormed into his office, and that was a very hatchet-faced looking secretary sitting there, and we said can we speak to Mr. So-and-So immediately. She said no, that he was busy, he's not here. People were going in and coming out and so on. He was clearly there. Eventually we gave up and we didn't get to him.

Barber:

Typical French!

Davies:

But there was more. At the same time this was happening, Ferranti had set up some other demonstration. The Paris air show was going on at the same time. They set up a demonstration using a satellite link. And the French law at that time said no satellite link could be operated except by the PTT. And this was only an experimental thing, so they argued with Ferranti, who said nonsense, we're going to — They were actually getting through the Frescotti database through this satellite link. But the satellite was owned by an international body, and so the French came along and changed the frequency of the satellite transponder overnight, in order to switch it off. And Ferranti discovered this, got new crystals sent out from Manchester, and they got back in operation! [Laughter] And these two things were happening at the same time. And as a result of this — I reported back and they reported back — diplomatic messages were sent from the foreign office to the French ambassador, and so on. They were really cross about it. Two beautiful examples of sabotage!

Louis Pouzin and PTT

Abbate:

I am interested in your relations with the French, even before then. Did you know about the Cyclades network? You were in contact with those people?

Davies:

Oh yes, very well. [Laughs] Once I was with Louis Pouzin, and we were doing, I am not sure what it was. Louis Pouzin worked for the government, but he was very cross with the PTT because they were always bearing down on him, and they told him that he must turn up on a certain day at a certain time. He must come, together with his colleagues, and explain to them exactly what we were doing about packet switching. It was a bit of an order to come at a certain time. And so we arranged to come, we were quite happy to come to Paris, and he said, "Well, the meeting is at 1:30 or something, let's go and have lunch beforehand at a very nice restaurant at 12:00." We thought, there's not enough time beforehand, but we sat down and had lunch. And the lunch went on and on, and he said, let's have another drink before we go. He was deliberately slowing us down until we were about an hour late. We went over there, us feeling sheepish, and people were going frantic. There were secretaries running in all directions, trying to find out where these people were who'd mucked up their schedule! It was all a ploy by Louis Pouzin to annoy them. The French PTT is extremely autocratic — and so were the Germans, the Germans were even worse. The German Bundespost, together with Siemens, they used to believe they ruled the world.

Barber:

GMD had a network that was entirely a circuit network, end-to-end. That was interesting. But that was part of the problem that I had with EIN, because you had some of the countries were convinced that there had to be circuit switching, and some of them that it had to be different.

Davies:

I went along to Siemens once and talked to them, and they actually used the words, they accused me of technical — they were really saying that I was being impertinent by suggesting anything like packet switching. I can't remember the exact words, but it amounted to that, that I was challenging the whole of their authority.

Barber:

Actually we got along fairly well with the French, particularly with Louis and Louis's group, Micharchien.

Davies:

The top levels of the PTT were very —

Abbate:

Well they weren't working for the PTT at first, were they?

Barber:

No, they never were. They were IRIA in those days, called INRIA now.

Davies:

It was afterwards that he was working for the PTT.

Barber:

Yes, quite a long time afterwards. It was Remi Dupres who had the RCP network, and Louis, and I really don't think that they liked one another anyway, quite apart from one being the PTT man and the other being in the ministry for industry.

Davies:

That's why Louis could afford to rile the PTT.

Barber:

Well, we were in the way! We didn't set out — we would have been reasonably fireproof if we had, I suppose, but we wouldn't have got very far.

Abbate:

So you were exchanging ideas?

Barber:

Yes, in fact Louie, his English was very good, and he did a lot of thinking about Internet too, but he called it "catenet," and in the end it became called Internet instead.

Abbate:

Was Hugo Zimmerman there then, or did he come later? I guess he was working on OSI.

Barber:

Hugo Zimmerman, oh yes. He was at that same place, at INRIA.

Abbate:

I know that they jointly wrote some papers.

Barber:

There was Micharchien and Hugo — formed a company. They made the operating system.

References

  1. The Alvey Directorate was set up in 1983 (for five years) to establish and manage a £300 million research support program involving industry, government labs and academics.
  2. Dr. David Yates, then headed the group that developed "Scrapbook." He later became one of Donald Davies' two deputies and so replaced Derek Barber (when he was seconded to head EIN)
  3. I was then his deputy superintendent (Division heads at NPL are called Superintendents)
  4. COST (Cooperation in Science and Technology) was a CEC research program. COST (Project)II — became the EIN project.