Originally aired on December 17, 2021 @ 10:00 AM - 10:30 AM EDT
Join Cloudflare CTO John Graham-Cumming for a conversation with
Dr. Ben Segal , early Internet pioneer and CERN's first official TCP/IP Coordinator, and Francois Fluckiger , director of the CERN School of Computing. Welcome to this special "30 years of the Web" segment on Cloudflare TV. I am John Graham-Cumming and I've got two esteemed guests from CERN who are going to talk to us about things that are happening, well actually probably 40 years ago rather than 30 years ago. Welcome, Francois. Welcome, Ben. Francois, do you want to kick off with an introduction? Yes, if you tell me what I should say. Well, let's put it this way. You did some things at CERN. What were you doing at CERN 30 years ago? So I was in charge of what we called external networking, which was a job to connect cells to the outside world. So I started the job in 1981, 40 years ago. At that time, we had two minuscule leased lines, one to Paris, another one to Clarendon in the UK. And my job has been to select the technologies to keep that, to expand those two connections to a certain level. And 30 years ago, we were in '91 from the two minuscule leased lines that we had ten years before. The CERN network has grown to such a level where it was the major hub of the Internet in Europe in '91, believe it or not, that 80% of the installed bandwidth for the Internet in Europe was organized as a big star. Around 80% was dominating in Geneva at CERN. And one of the reasons why this had been expanded so far, there were two reasons. One, that, but Ben will comment on that, we believe we had, by '88 to '89, selected the right technology to build that network, the Internet technology, more exactly the TCP/IP technology. And the second we started to have operated one new application to run on top of TCP/IP. It had just been proposed two years before in '89 by a young engineer called Tim Berners-Lee. And by '91, '92, 30 years ago, there were already pages being served by the first ever web server connected just at the center of the European Internet hub in Europe. All right, Ben, so why don't you take us through your part in this? So you were overlapping with Francois at CERN? Okay. What were you up to? Well, I came to CERN 50 years ago in '71, long before we had these sorts of problems. We didn't have things like networks. Actually, we have some rudimentary networks. By the way, that's simultaneous with the start of the ARPAnet. ARPAnet started in '69. Okay. So I'm as old, my career at CERN is as old as the ARPANET, which became the Internet. Until '83, I didn't work on general networking problems. I worked on satellite connections, I worked on some other things at CERN. But in '83 I started to work in a little section called Distributed Computing with a boss who became and stayed my boss the whole time after that, who I call Leslie Robertson, without whom none of this would have happened at CERN anyway. So, yeah. So we started to look for protocols that would help us to connect computers together. And we had a lot of different sorts of computers at CERN like most people did. We had the big mainframes, IBM, we had a lot of the DEC machines, Digital Equipment, they've disappeared and so on. And all this stuff, a lot of this stuff we wanted to share data on and it was very primitive. In the seventies, CERN designed its own network and connected a few of these things together. It was called CERNET, but it only allowed file transfer and in a couple of machines it allowed log on, remote log on, but that was it. So we came up with this TCP/IP story and we found that we could actually buy or get for free software to put on most of these systems that we wanted to test it on. And then we ran into this difficulty that there were forces against using TCP/IP. In particular, it was American. CERN is European. It was what we call today disruptive. It was cheap and sort of worked, but it was competing with the standard solutions and in particular with the monopolistic solutions which were being imposed by the telecoms in Europe and elsewhere. In fact, it was explicitly opposed by the PTTs. I remember the German PTT, of which everyone was terrified, they were supposed to be even looking to see if people were using TCP/IP and doing bad things and this was actually not true at all, as we found out later. So we had this climate of fear and discouragement coming from the management, but they did allow me, in the end, to coordinate the introduction within the lab. And so we got these things running within the lab. In those days, they run quite slowly on those systems, so the software wasn't very good, but it worked and the physicists started to be interested. And it wasn't until '88, end of '88, beginning of '89, that CERN managerially accepted these protocols in spite of their controversial nature and went with it. And at that point, I stepped out of the picture and I had another big project on my hand, which was getting rid of mainframes and replacing mainframes by clusters of small, powerful machines. And so I didn't play any part later in the TCP/IP explosion, which happened at CERN later, which Francois can talk about. Finally, I met Tim Berners-Lee, who is the real subject of this conversation, in, let's see, that would have been about '86, '87 when he was having these ideas about what became the web and he heard about TCP/IP and he came to me for coaching, actually. I taught him Internet programing and he incorporated that, of course, into his invention. And after that, I mean, the world changed. But I just want to make this point, that we thought we were trying to solve technical problems. And in the end, we were running into serious political issues. And to this day, of course, the Internet, which we as innocent pioneers, we are unfit for now. I mean, there are social issues of such gravity that it's beyond everyone. We thought we would connect everyone. We dreamed we would connect everyone. But in fact, we connected the bad stuff as well as the good stuff and you know, and so it's still an unwritten book in front of us. Before we get to this, the Tim Berners-Lee and the web part of it, I want to wind back, though, to the beginning of this, because in 1981 is when you would get RFC 793, which is the standardization of TCP and then this is where you're going to kick... And then it, it was written in '74 and it became yeah, they switched. It sort of became this big thing. But what... you were doing networking before this with your own thing. But, but the telcos also had something, right? There was X.25 was a really big deal in Europe. So let's talk about X.25 because I remember, I'm old enough to remember using X.25 systems and they were ,they seemed pretty exciting. You could connect to things and you could discover things on the network. What happened? Why don't we use X.25 now? Well, let Francois talk about that. Yeah, that's difficult for me to talk about X.25 because I am one of the inventors of the technology back in '73. It's a great technology. I liked it very much, but after ten years it was clear to me and to all my colleagues at CERN and elsewhere that this was an outdated technology. The principle of X.25 is that it has been designed by the telecommunication operators, by the telcos to mimic the mechanisms, the particulars, the charging mechanisms of the telephone. So X.25 is based on the principle that you have calls. No computer can talk to another one without making a call. The duration of that call is recorded, the traffic of that call is also recorded and you pay per call and for your traffic and for the duration. This is a nightmare. Your own charge, as it was my case of of networking in a given organization. You never know at the beginning of the year how much you need to plan for the expense of your communication costs. And more importantly, and this is why the web would have never worked with something like X.25. You need to place a call before you do anything, which means that, let's assume... the web mechanism. You click on a link. Before the first packet is sent to make the TCP connection, you need to set up a call, a call from your laptop to the remote server and back, a minimum of 400 milliseconds. It would never work. Each time you click on a link, you'd have to set up that code, so it would not work. So maybe a word about the political climate at that time because Ben mentioned it. Certainly what we have to keep in mind were the servers in Europe. And from '86 to '94, there was, in Europe, the protocol war was raging at that time. It's not such a small element in that battle, because all the governments in Europe, except a few of them, I will give their names because they deserve to be to be mentioned, those countries. But except a few countries, a handful of countries, all the governments and the European Commission had decided that we needed technologies and protocol which were vendor independent, which were not proprietary, which was good, but which was also technologies where Europe as a sort of age, ahead of the Americans. So they adopted as a strategic direction this oversized suite of protocols, including X.25. And the whole of Europe was pushing that way. And when Ben and others came and said, okay, well there is another non-proprietary technology, but that works because well, after a while we'd convinced ourselves this OSI would never work. The rest of Europe was against us, and for us at CERN, it was difficult to resist because we are paid by governments in Europe. And I can tell you that there were pressures and there were direct threats from governments or ministries of industry in certain countries to prevent us to use this TCP/IP. Fortunately, what helped us is the fact that we are mission-oriented. What we say is Our mission is science. Our mission is to find new particles, is physics. Our mission is not to help European industry, even though our citizens would like it, would love it. So please remember that we have to select technology that helps our mission. And this is why we were allowed by the governments to carry on with TCP/IP from '88, '89. We were not alone, but we were only a few of organizations. The Scandinavian countries. Sweden. Denmark. Finland, Norway. The Netherlands. Ireland, Austria and CERN and that's all. And in '91, these countries plus CERN, we did set up the first IP backbone in Europe. We called it EBONE but you see that the rest of Europe was against us to set up the whole picture. Now, a final word on that. One of the reasons why the invention has worked in particular its transport mechanism to transport from one place to another one, http, it has worked because it is a connection-less technology. You do not need to make a connection between the two computers through a network and it is the same spirit as IP. The mechanism by which the web works, the web server says, Take a request. Serve it. Forget it. Take a request. Serve it. Forget it. This is how a server works. It is the same mechanism as an IP router. An IP router takes a packet, forwards it, forgets it. Take a packet, forwards it, forgets it. The two are fully compatible because they have the same philosophy. And this is sort of an ironic thing, right, which is that... so European countries were trying to hang onto X.25 with the connection-oriented kind of thing, which would have prevented the creation of the web, which actually was sort of funny. It was like by trying to not be American, they would have actually ended up in the front. What happened was by adopting this sort of, as you say, disruptive technology, slightly worse, not not quite so finished technology, they actually ended up creating an environment where something else gets made. Ben said rightly that this famous waist-high pile of of protocols would have never worked because it was by far too complicated, unlike IP, which is extremely simple. And even though it worked, this would have given less to us as the Americans, which had much stronger companies. And this is, on the contrary, like technology, like IP, that anyone, anyone knowing how to build that switch knows how to build to construct an IP switch, which is much simpler. So it was precisely TCP/IP, which was well stood up to American industry allies this gas factory that was, I was at the time, which never worked again. Ben, you were going to say something? Yeah, actually the European Union was able to save face because although they had fought against these cancerous American protocols for so long when the Web came out and became a worldwide success, they learned that it was invented in Geneva, and that's in Europe, more or less, I should say. By a British person, more or less. It enabled them to save face and then get behind the Web. Right. Right. And I wanted to go back to something you said Ben, which is you taught Tim Berners-Lee Internet programming. So what did what did Internet programing look like at this point? This is Berkeley Sockets model? Absolutely, absolutely. Well done. And you see the other protocols usually didn't have a standard interface. That was so beautiful about the... as long as the thing had Berkeley Unix on it or an equivalent, you could write software running across heterogeneous systems. And it was it was wonderful. And that's the whole... by the way, the fact that the TCP/IP protocols were incorporated into UNIX in '83 by Bill Joy and company in Berkeley... in California in Berkeley. That was another big reason why they succeeded, because they penetrated like that all universities, all the research institutes that were using, starting to use Unix like we did and you got TCP/IP for free, whereas you didn't get... Yes, so socket programing. Of course, Tim learned that in 10 minutes, but at least he had it. Okay. Yeah. I mean, it gives you a layer which you can then communicate with another machine. Presumably, what other machines was his machine talking to within CERN? Were they, were they going over Ethernet to some other devices? The very first thing, one of the very first things that Tim did was connect, and you'll laugh, to the telephone system, the telephone book system. So you could look up a number on the certain telephone system via the web, which was, funnily enough, one of the reasons why it attracted a lot of attention because there wasn't anything out on the web when his first server came up. It was the only thing there. But very soon, when Mosaic came along, you know, a beautiful, sexy browser with with color and everything running on PCs, then the web really made a hit and you could visit the Louvre and this sort of thing. You remember those beautiful sites? Yeah. Yeah. So but that's another story. Did anyone from a telco see the irony that the first thing that they made available was a telephone directory on the web? I mean... Oh, well no, it was under the CERN telephone. But it must have been... you could call someone, you could set up a call with someone using this technology. As you said to us just before we started, the French had Minitel, which had many of the characteristics of the web. It just didn't have international standards that were going to work. It had a French keyboard to begin with. That's enough to kill off most. I have to tell you, growing up in Britain at that time, I was so jealous of Minitel. I would read about it in magazines in England and I'd be like, Oh my goodness, you can get a terminal in your home. I was so jealous I actually bought one. I have one right next to me here. I fulfilled it like a 30-year-old wish that I'd have my own Minitel. They gave it away for free instead of the phone book. It was marvelous. Yeah, yeah. But the British did have a thing. I've forgotten the name. Yes, I know. Prestel. And believe me, it was not as good as Minitel. It had this weird setup. The beauty of Minitel, with a little bit of Minitel reminiscence, is that it's one device that you plug in, has a keyboard, has everything. Prestel was a weird thing with an acoustic coupler and some combination of stuff. It was sort of a bit Heath Robinson, whereas the French sort of dirigiste kind of thing was like, Right, we'll just make a million of these and give them away. Well, the whole world was like that at that time. This zoo of different protocols and standards and hardware. It was a miracle what Tim did. He used to say, we just need to agree on a few simple things. And he specified those simple things, just three simple things. And he solved not just the networking problem, he solved the data presentation problem. Right? Right? Fantastic. Yes. Yes. Yeah. Yeah. Go ahead, Francois. Ben said something important. He actually specified, but not only specifying. One tendency we tend to have sometimes in Europe is to to say how things should work, another way to do specifications, to specify mechanisms, and then stop here and say, now it's to someone else to implement, and to maintain once you implement. At most, you may make a prototype and then you forget it. So Tim not only did the full specification, but implementation. And this is, this was not that common in people, when people do specify start-ups, particularly in telecommunications. Maybe another reason... I said sometimes that the advent of the Web is a mix of chance and necessity. In fact, this is Jacques Monod, the Nobel Prize laureate, who said "Life is a result of chance and necessity." So there were a lot of chance for the Web. The little of necessity. One of the chances was that it came to simplify was needing to technologies. One, to describe the format of the documents. And this is why we developed HTML and it needed the mechanism to transfer these pages from one computer to another one. And this is why he designed HTTP. Now, it turned out that the chance was that at CERN, for the first problem, the document format, there were, at that time they were there, some of the best document format experts in the world at CERN. We had been a key, CERN had been key in the development of specification of SGML at that time was one of the standouts. So you could discuss with some of the best experts. And when it came to him to look for a technology to support, a model to design this HTTP, he had again the chance to have, at CERN, some of the best experts, in particular, one of the best experts, Ben, on IP. So it was a chance that he you could rely on world-class expertise in these two domains. Not only could he rely on those people, but he managed to become himself a world-class expert in IP and in document format. Right. Right. So think, these two things are quite distant. I don't know who has, who masters thoroughly such two quite distinct technologies. This is why the two work so well together, HTTP and HTML, because they come from the same brain. You make a really good point about that. Standardization, kind of like X.25 was a great example. I remember coveting having the standard so I could read the standards and then you get... he just implements the whole thing and it just works. There's something in here, not just about that, but about the openness of all this stuff. The, you know, the Unix, you can see the source code. The protocols were there out, defined. And then, you know this better than me, afterwards. You know, this decision to public domain it all and give the software away that really accelerated this in a way that I feel like these proprietary or very carefully controlled standards really holds back technologies. That's absolutely true. Tim, apart from being a genius in picking the right elements for a standard, he picked what he needed and not too much. He knew about open source. Most people didn't know about open source technologies in those days at CERN. And he put out the software on, he couldn't put it out on the web, but he put it out on Usenet, where the bulletin boards and his boss, Mike Sendall, whose name has to be mentioned, a wonderful man who gave Tim the freedom to do this. He, when Tim went to Mike Sendall and asked him, I'd like to put this software out because we need help, we need programmers and we can't get them at CERN. How do I do it? And Mike said to him, Well, officially you should go to the legal service, but they will wait a long time. They'll look at it for a long time, and they may say no, in which case you can't do it. And they're very unlikely to say yes, so just do it. Yeah. But yeah, on that very point, there is a story which is not very well-known... But first, before going any further, there are sometimes I would say, wrong messages which we may hear here and there such that Tim Berners-Lee was generous enough to not make money by selling his web, but by putting it in the public, in the public domain. But it has never been an option. When Ben and I joined CERN, when Tim Berners-Lee joined CERN, we knew, when signing our contract with CERN that we would be paid by the European citizens, but all that we do - would belong to the American citizens,to the European citizens and not to us. In other words, unlike many universities or even research centers, we cannot take patents. We do not own anything. All that we do belongs to CERN. So there was no option for Tim to be reselling to anyone the web. The web belonged, as always, belonged to CERN. Now it turns out that in March '93, CERN convinced the CERN management, or more exactly its legal service at that time at CERN, that the good thing to make sure that all the web would be freely available to anyone was to put it in the public domain. And CERN decided to put the Web in the public domain. Oh, yeah. Yeah. However, it had been a mistake, which could have had very serious consequences. When Tim left CERN in July of '94, a week later I was asked to go work on his technical team, in addition to my other job. So I took the job and it was a time where the team, the Web team at CERN had to release the long-awaited version free of of the Web software. So I looked at it and I remembered this public domain story, even though I was not involved at that time. And then I realized what it meant, because public domain means... the idea is not that. Public domain means, I have this object, I want everyone to use it and to be able to modify it freely. So I say I put it in the public domain, which means I relinquish the intellectual property. This object no longer belongs to me, so please use it because it's even not to me a right. Right. If you do so, you run the risk that anyone say, Are you sure it's not yours? Okay. So I take it. I correct the bug. You need to change one small, one tiny thing. I correct the bug and now it's mine and I deny anyone to use it free. So this is a risk of appropriation. This should not have been done. And no one does that ever for software. Right. So I rushed to the legal service. I said, you know what has been done 8 months ago? And they say no. Oh, my God, who has done that? Oh, the guy had left CERN from that time. We rushed to the IPO, Intellectual Property Office, in Geneva and say, okay, what should we do? And they said to me, Well, when is your next release? I said, Version three, I should release it about two months from now, three months. Do it now. Do it immediately. Do not wait a single day. And the week after we released it. But on the open source, which is something that Tim ignored. Tim did not know the details at that time of the legality of open source. What he knew at that time was a spirit of collective work, of an open source, involving people from, young people and students from university, to help him working, working. But the legality, he did not know. So what we did is that. Okay, look, guys, this object now is version three. It still belongs to CERN. We own it. So it's released under open source. We own it. No one can take it, but we give it to everyone, to anyone, irrespective of what you are, the irrevocable and perpetual right to use it freely. But no one can take it. This is open source. And later on the version four and the others done by MIT by Tim Berners-Lee at MIT, they used the same mechanism and no longer anyone used the public domain, but still there were 8 months during which someone could have taken the Web. On that note, I'm going to thank you both for spending half an hour talking to me about networking stuff because I really like networking a bit too much for my own good. And we managed to avoid talking about token ring. Otherwise we would have spent a long time talking about my great love of token ring. Thank you both for your part in getting TCP/IP going in Europe. That's how I got on the Internet in the eighties. And for your part in getting the web going, we're all, we're all living this this web experience together. I wish you both a good evening. Thank you. Thank you.