A Brief History of Unix and the Internet

Includes the following content, audio and video links available at top of page:

Original TechNetCast interview with Peter Salus. Available in video and audio. Parts of the transcript of this interview are included below.

A Brief History of Unix and the Internet Peter Salus presentation session at USENIX 1999, in audio. Q&A; session includes interventions from Dennis Ritchie, co-creator of Unix, and other pioneers.

Peter Salus is the author of A Quarter Century of UNIX (Addison-Wesley, 1994) and Casting the Net (Addison-Wesley, 1994). He is also author of numerous articles and the editor-in-chief of the Handbook of Programming Languages (Macmillan). Peter served as Vice President of the Free Software Foundation and Director of the Tcl/Tk Consortium.

More TechNetCast At USENIX 1999 programs.

DDJ: Your book, A Quarter Century of Unix (Addison-Wesley, 1994), is now five years old. The Unix world has gone through several transformations since then, most notably with the momentum behind Linux.

PS: Linux already existed when I wrote A Quarter Century. There are a couple of pages on Linux towards the end of the book. We have seen a development from roughly speaking one user, Linus himself in 1991, to ten users a year later, to somewhere in the 7-10 million range right now. That's quite stupendous.

DDJ: From your perspective, what explains that growth?

PS: I happen to be a Linux user myself. I don't mind admitting this --even in public. Up to the very early '90s there wasn't a good version of Unix that was usable on the 386, the 486 and now the various Pentium chips. There was one version that was used as a teaching Unix, and that was Andy Tannenbaum's Minix

DDJ: The Unix that Andy Tannenbaum wrote from scratch.

PS: That's right. In 1978 or '79 in order to get around the incredible licensing. If you pardon me I can come back in a moment to that because we're nearing the significant anniversary for what I would call American corporate insanity. At any rate, Linus was a computer science student in Finland. He sat down and wrote this from scratch. It's clearly a clone. It's clearly based on Unix. It employs all the Unix applications and tools and all of the current things, and even those done in the last few years. It's POSIX-compliant. POSIX is P1003.1, IEEE's standard for inter-operability and portability of operating systems. There's a footnote for you... Linux is a POSIX compliant kernel with a lot of stuff, mostly Unix and Free Software Foundation materials, around it --tools and applications. It will work on everything from an old 286 box up. You can run it with as little as 4 MB and a couple hundred MB of storage. In other words, practically any old 286 box would actually enable you to boot and any 486 or Pentium box enable you to run in a good fashion. And you get the source.

DDJ: So the aggregation of all these features -Unix, runs on Intel, open-source- is what propelled Linux. And no other Unix OS had all these features at the time.

PS: Yes, and Linux then got this vast community of enthusiasts, and I will call them enthusiasts, who rewrote tools, made things better, reincorporated things. [And this is] something which comes right out of the Unix tradition.

DDJ: It seems that many of these new Linux users were also new to Unix. They were not previously Unix users.

PS: Yes, but that was always true. I consider myself an old man... We're now generating ever larger numbers of young people who are familiar with computers. I took my first computer course at IBM, which had its labs near Columbia University in Manhattan at that time, in May-June of 1957. What I studied in 1957 bears no resemblance whatsoever to any of the things I've used in '77 or '87 or '97. But I've just gotten good at relearning new stuff. Most of the people you will see [at Usenix] are well under forty. They've been brought up with computers. They've been brought up thinking they can do things, and they're really interested in something that is practical, compact, easy to handle, easy to attach things to, and Linux in all of its versions allows that. I mentioned the source availability. This is really important. There's nothing you can do about a Microsoft product. But if I don't like something in Linux, I've got the source there on my disk. I can go in there, I can change it. In fact I can change it enough so it breaks my machine. But that's all right because I can slide it out, redo it then from my CD-ROM, and I'm back where I was. I think that people want to be able to control their environments and their fates. Therefore, "I want to be able to use X and Y and Z and I don't want W" is an important thing. Not "you've got to buy the entire thing."

DDJ: In the early days Unix was an open system, and that contributed to its growth. Then, somehow, along the way, something happened.

PS: That's the anniversary [I was referring to]... There were basically two streams in Unix. One is the BSD, the Berkeley software stream, and the other one is the AT&T; stream. In [about] 1978, Berkeley software offered its first tape of software. All you had to do was send in sixty bucks and you got a spool of tape with the bits on it in the return mail. At the Toronto Usenix meeting of June 1979, the gentleman who represented licensing for Western Electric, a wholly owned branch of AT&T;, arrived with a slide and with information for all the members. The information was that the price for the licensing was going up to between $20,000-$40,000 per CPU. There was dismay throughout the population. I think it was one of the great mistakes that AT&T; ever made because they didn't realize what they had until they put it up to this crazy price. By doing so they forced the users to get together and cooperate with each other. Usenix was formed as a Unix users group because AT&T; sold ("gave away") Unix but with no bug fixes and no support. And you had to pay in advance. Now this meant that if something didn't work right or if you couldn't figure it out, you had no way to turn except to the other users. So as soon as there were two or three dozen users, [they] got together at what I would call the first Usenix meeting, in May 1974.

DDJ: Then start the dark ages of Unix. Unix becomes fragmented, proprietary.

PS: Proprietary. Absolutely. What happened at this point is that with these licensings, AT&T; also went to commercial instead of just going in for the research and academic crowd. One company, "Interactive Systems" (they were bought up by somebody a couple years ago) [came out with] what I would call the first commercial version of Unix. But soon there was a relative proliferation. The '80s saw a very large number of "Unices". Most of these either grew out of what was going on in New Jersey at AT&T; or grew out of what was going on at Berkeley. But there were incredible sets of interactions. I don't know whether your users will recognize vi... vi is the standard editor, the vi stands for visual because it was the visual editor. The original Unix editor was a horrible line editor called ed for editor. I mean, there we go, another intellectual sort of thing... ed is beloved only by people who every so often have to actually boot a Unix machine where everything else has gone kaflooey... It really does work. But it's not a very nice job. George Coulouris, who was then at Queen Mary College at the University of London, got a copy of this from the labs and hated it and changed it into an editor which was called em, which stood for "ed for mortals". The next year he went to the University of California-Berkeley on sabbatical, which academics do periodically. While he was there he was running it and, as he put it, this wild kid of a graduate student saw it, asked about it and made a copy of it. George went off to the labs for two weeks to talk to the people in New Jersey. When he came back,this kid had turned it into ex, another line editor. The kid of course was Bill Joy, who then went on to become one of the founders of Sun, and who in the next revision turned it into vi. vi got incorporated into the AT&T; version of the Unix distribution. What you have then is something that comes out of Bell Labs and goes to the University of London. The University of London revision of it goes to Berkeley. Some kid at Berkeley revises it and then that gets incorporated back in the next release. That's about as open and cooperative a system that you could have.

DDJ: And we're now seeing similar interactions in the Linux community.

PS: Exactly. The more things change, the more they stay the same. It's a wonderful community. I walk around here and I really like these guys. They're smart, they're able and they're interested in making the system work well.

DDJ: Rather than making money

PS: None of them is averse to earning money. I'm not averse to earning money though I'll never be a multi-millionaire in Silicon Valley

DDJ: Bill Joy is

PS: Luckily, I'd say... He's lucky. Other people do it with a nice head start and a little bit of a push, this sort of thing.

DDJ: We talked about fragmentation. Are we seeing now a convergence of Unix variants?

PS: Well, the Unix variants have been disappearing, let's put it that way. Siemens in Germany had a system called Sinix, IBM had AIX, Apple had AUX, HP had HPUX, the version of Unix where everyone was really glad they weren't the Packard-Hewlett company... But all of these have effectively gone away. Sun went from Sun OS to Solaris, which is an AT&T-added-on-to; Berkeley distribution. BSDI is still running pretty much straight Berkeley distribution sorts of output. But you mustn't think that because there are variants they're not clear cut. If I sit down at a machine that I know is a Unix machine, it doesn't really matter what variant it's using. It's got almost all the commands, and the applications are going to be absolutely identical for me.

DDJ: In some ways, the Internet is an example of this interoperability. The reason machines all over the Internet were interoperable was that they all used TCP and UDP/IP.

PS: That's right. But it's even more than that. The original Internet --the incredible, wonderful 1969 investment by our government of barely over a million dollars, probably the best thing that government research has ever invested in-- was [built] to connect four different university research sites that were operating four different machines using four different operating systems. That was the real battle. We had e-mail, for example, between two users on the same machine before that. But the trick was to do it heterogeneously. Just about the time Unix came out of the closet in '74 or '75, some guys at the University of Illinois, Steve Holmgren was the lead, posted on the early Internet an RFC referring to network Unix. It's RFC 651 or something like that. The point was, "we have a good, compact, flexible operating system that we can use [as a network OS] because we can use homogenous commands across three dozen machines around the world. It's very hard for us to come to grips with just what the growth of use and geographic growth [of the Internet] has been like. We've gone from four machines --UCLA, University of California-Santa Barbara, the Stanford Research Institute and University of Utah-- in December, 1969, to, roughly speaking, 43 million hosts in January, 1999.

DDJ: And In those early days no one could have predicted that this network would be used for the type of applications it is used for today

PS: There were two applications at the very beginning rlogin to enable logins into remote machines and FTP [that made it possible to] transfer files from one machine to another. The next big deal came a year and a half later when Ray Tomlinson invented e-mail. The interesting thing about Ray inventing e-mail is that within a year mail among the various hosts was the largest use. And these were not dumb guys. These were really smart guys, the graduate students and the supervisors that designed it.

DDJ: Was email as much the result of a careful design process as much as a quick, practical solution to a pressing problem?

PS: What Ray did was put effectively a pair of saddlebags on the pony express that was FTP. They were special files that you were transferring from one place to another. The sort of things we have now in sendmail are ten years later down the road. But the thing is that you need these developments. Mail became an incredibly fine thing. The next real big explosion, I think, came with the Usenet newsgroups. It started out with a couple of punky mailing lists, SF-LOVERS, UNIX-BUGS. Very straightforward. Two institutions in North Carolina were the original originators of the sites Jim Ellis and Tom Truscott at Duke and Steve Bellovin at the University of North Carolina [put it together]. Ellis came to the Usenix meeting at the University of Delaware and told people about it, and that was it. Soon there were hundreds of groups. Now there are tens of thousands of groups.

DDJ: As anyone who has tried to download a Usenet newsgroup list over a slow connection knows

PS: We've always been lousy at predictions. So I have a totally opaque crystal ball [when it comes to] saying what I think will happen in the next decade. I really mean that. If you had asked me in the '60s, I would have told you that the big iron would be around forever and that the software would be the stuff that got swapped in and out. Well, I would have been wrong. Who would have foreseen the PC, for example. And we would not have a Y2K problem if it wasn't for the fact that 40 year old and 30 year old software is still in use so that we have these legacies. That's exactly the opposite of what I think anyone would have told you in those days.

DDJ: One clear trend is that Internet is becoming more and more synonymous with the Web, and HTTP is becoming the protocol of the Internet, in some ways the transport protocol of the Internet.

PS: Right. I think it's true but I don't think it's ever going to be all [HTTP]. Don't forget that in the course of the last five or six years BITNET has disappeared completely and that FidoNet has died, although it is still clinging in certain corners of non-North America. UUCP links are unheard of. Yet I can remember in the '70s when UUCP (Unix to Unix Copy Protocol) was just absolutely wonderful and the way to move things around. We don't have to do that anymore. IPv6 is really going to be an impressive cut forward, or at least I think so. I love it.

TechNetCast Catalog:
usenix technical conference internet unix Peter Salus 

Related Programs:
• Bill Joy on Openess and Community
• What's Wrong with HTTP and Why It Doesn't Matter
• Eric Allman
• 1999 USENIX Technical Conference

FORUM

add a message to this thread

1999 USENIX Technical Conference
posted by Chris Worley 1999-06-14 [#35]

You taped Linus' speech. When will it show?
[reply]

1999 USENIX Technical Conference
posted by lamefrog [member]1999-06-14 [#36]

The Linux BOF -video and transcript- will be released
Monday, June 21.

[reply]

Linus Torvalds at USENIX Linux BOF
posted by lamefrog 1999-06-18 [#38]

[4 replies, expand] [reply]

Linux on a 286?!?
posted by Rhys Dyfrgi 1999-07-21 [#192]

Linux won't run on a 286, it's a 32-bit O/S, and requires at least a 386.
[2 replies, expand] [reply]