Chapter 09 - Networking: Emergence 1979-1981
9.6 Zilog
Ralph Ungermann, one of the two founders of Zilog, recruited Bass from the faculty of the University of California at Berkeley to run the software group, which, along with the hardware and semiconductor groups, reported directly to him. Bass in turn hired John Davidson from BBN, his buddy from the ALOHAnet days, and Joe Kennedy. In 1976, Bass hired Judith Estrin, a graduate of Stanford well grounded in TCP, having worked with Cerf, and very familiar with Ethernet. Even with this considerable talent, however, the impetus to start an internal networking project came from a very unexpected source.
Bruce Hunt felt very lucky when Kennedy hired him in May 1977. Only a year earlier, he had been a programmer at the Stanford Linear Accelerator (SLAC), when, in one of those casual conversations that leads to the unexpected, Hunt remembers Joe Wells, the Director of the Stanford Center for Information Processing, asking him:
‘Bruce, do you know how much it costs to connect a terminal to our computer system?’ that was an IBM 370 system at the time. I said: ‘No.’ He said: ‘About $3,000,’ and I thought that was astonishing, since a terminal cost $2,000. So I started to really think about how one might deal with the issue. The thing that was interesting was that in the July 1976 issue of the ACM, there appeared Bob Metcalfe’s famous Ethernet paper, which I read, and went: ‘Wow, that’s really amazing!’ So I started thinking about that, and just about that time, Bernard Puteau, who was working on the Z-8000 microprocessor at Zilog, brought by a copy of the first Zilog SIO specification sheet. I looked at that and said: ‘You put these things together and you could actually create a network’ I then spent a huge number of evenings just sitting down and thinking about the issues. Thinking about how you could construct the hardware and so on. I essentially came up with the basic scheme of how it could be done.
In landing a job with Zilog, Hunt had the opportunity to see if his ideas would work:
I went down to Zilog and literally built what was called Ariel over that summer with a guy named Ben Laws. Ben had come from Xerox PARC. Ben was a super guy to work with. I explained to him what I was up to and he said: ‘Oh yeah, there are some great chips that we could use.’ So we sat down and he did the hardware and looked at my simple designs and turned them into something that would work, and then we built the stuff. I did all the basic software that made it all work. We had the thing up by the end of the summer. In fact, as I recall, I started spending about 60% of my time doing nothing but giving demos of how the system would work, because it really ran.
Hunt’s Ariel project then got sucked up into the energy and chaos of Zilog with “some very heated discussions” of Ariel and its implications. Ungermann recalls: “We had some yelling and screaming matches where I was convinced that we had to have a low-cost, semiconductor-based interconnect strategy.” And an interconnect strategy was only one of many projects, underway or proposed. As Ungermann remembers:
We were a company that was encouraged by our parent to go in as many directions as possible, because we were the most prolific producers of products. So they were constantly saying: ‘Can’t you do this? Can’t you do that?’ So as a company, we went in ten different directions. We had stuff sprouting out all over the place! I can remember we had a board meeting and we walked Exxon around once to see the technology that we had going. I remember coming back to my office and thinking: ‘My God, unbelievable! It was unbelievable.’ We had everything in the world. We had very, very good people, and we did lots of things, but we didn’t make lots of money. My sense of Xerox PARC was that there was some kind of overall systems design which they were working towards. That may be oversimplifying it, but my sense of it was that Zilog had as much talent and as much energy, but it was more dispersed and it didn’t have any mission or objective or unifying force to it.
Characteristic of the Zilog culture were the frequent parties held by Estrin and her roommate, Barbara Koalkin, who worked at PARC. These parties facilitated a cross-pollination of the PARC/Zilog cultures and activities. Invariably, small groups huddled intensely for hours discussing new technologies just becoming available or even still just thoughts. Hunt remembers one party:
One of the people Judy had invited to one of her parties was Fouad Tobaji. Fouad had just accepted the job at Stanford as assistant professor of electrical engineering. Judy introduced us and said: ‘Fouad, you ought to know Bruce. Bruce has built this little network’ and Fouad says; ‘Wow, you know, I studied networks. That’s what I did for my PhD for Leonard Kleinrock. You ought to go look at my papers.’ So I did. I went and read all of his papers and his Ph.D. thesis and he had proposed a way to analyze networks. This was similar to Metcalfe but Metcalfe didn’t have the basic mathematical sophistication to carry it through. Well, Fouad had and he carried it through for the carrier sense multiple access scheme. In fact, he wrote four very famous papers. Well, when I read his papers, I said: ‘Gee, I can extend this to the collision detection case.’ So I did the derivations and added the extension of being able to handle collision detection and I took it to Fouad. Literally, I think I did it in about 12 days. After you read this stuff, your head is full, so you’ve got to dump it. Fouad said: ‘This is great Bruce. Let’s write a paper.’ Fouad later said: ‘Well, why don’t you think about coming back to Stanford and working as a graduate student.’ So I did.
In October 1978, Hunt left Zilog for Stanford, ostensibly as a leave of absence. Estrin became the new project leader for Ariel. Rosenthal invited Hunt and Tobaji to present a paper at the May Symposium.