An Interview with Watts Humphrey, Part 14: Jay Forrester, Program Pricing, and the FS System
This interview was provided courtesy of the Computer History Museum.
Another Boston Story
Humphrey: Right. When I was in
Booch: If I can clarify, what year would that have been?
Humphrey: That would have been
in about 1956-1957, along in there. This is actually in 1959. A little bit
later, they had me run for president of the
Barbara Humphrey: Dudley Buck was the inventor of core memory. His picture was on--
Humphrey: No, no. Dudley Buck was cryogenics. That was very cold computing or low temperature computing. But anyway, so I went over and had lunch with Jay Forrester. I frankly don’t remember much about the lunch. He was a very pleasant gentleman. But just a minor point that I happened to meet him was kind of interesting. I’ve been exposed to lots of people.
Booch: Remarkable. By the way, speaking of the Whirlwind, have you been out to the museum in the West Coast and seen what pieces of the Whirlwind they have?
Humphrey: I did. I did stop
off there. I mentioned something we had, which unfortunately we didn’t keep. When
Barbara was there at the Whirlwind, they were just converting from the old --
they had CRT tube memories. The old memory on the Whirlwind stored dots on -- I
guess it was charges -- on the Williams tubes and they scanned them. They had
to keep rescanning them to keep the memory. So they had this enormous
The Program Pricing Flap
Another thing I don’t
think I mentioned was the ACM meeting. This was in about 1967. It was before we
priced programming. Everybody thought that programming pricing was going to
come. There was lots going on in the subject. I was
invited to sit on a panel discussion on it at an ACM meeting in
Humphrey: Hillary Faw was the pricing guy on the FAA and later. He was then the Director of Policy Development on corporate, a job I later had. So I called Hillary and I told him what I was going to do, that I had put together this statement, and could I have him look at it and was it okay? So I did. In fact, I sent it to him. He looked at it and he called me back and said, “Yes, that’s fine. You go right ahead.” So when I got over to the meeting where the questions came up about pricing programming. I said, “Well, there’s a real misunderstanding here. I won’t tell you whether we’re going to price it or not, because that’s a policy decision the IBM company will make, but I should tell you that there’s a lot of misinformation on the subject.”
People think that if IBM priced its programs, there would be an enormous reduction in hardware costs. The estimates that ranged in the press were in the 40 to 50 percent range. If you want to figure out really what it costs and what might happen to the price based on cost (I can’t tell you, of course, how we’d price it), but based on cost, there’s no way you’d get a cut of more than maybe two or three percent in the hardware costs. The reason is because the way you price stuff covers costs. If you look at the costs for IBM and take out the marketing people and all of that, but you look at the general cost, what percentage of the IBM population are programmers? It’s less than three percent. So 40 percent is crazy. There’s no way you’d do that.
“The costs aren’t going to come down any more than about three percent.” This hit the press. When I got home, I had all these calls from IBM management. “What got you to you say that?” I told them it had been approved. I reviewed it with Hillary Faw and they were really upset that I hadn’t reviewed it with my boss, George Kennard. Straight nonsense. I didn’t say it to anybody, but I think people realized it. George is a prince of a guy, but he wouldn’t have said, “No, don’t say it.” He’d have had no problem with it. But in any event, I did get by without a wrist slap coming. But I thought that was kind of interesting, because people had a wholly unrealistic view of what might happen. In fact, when we did price programming, the hardware price was actually reduced by 3%.
Booch: By the way, you mentioned that the programmer population was four or so percent. What was the entire IBM population around that time? Do you have a number?
Humphrey: I’m afraid I don’t remember. I know when I retired in 1986, it was 420,000 people.
Booch: I do recall the IBM archivists have that information, so we can dig up the numbers. But I was just curious as to four percent of what. I’ll see if I can find that number.
The FS System
Humphrey: Yes. It wasn’t a big number. The FS system -- let me talk about that. In 1970, after Bob Evans became division president, a fellow named George Raden, a real bright guy, had been in programming. He’d gone into research. I think it was George. I know he was involved in it, but I’m not sure if he was the guy that actually invented it or not. But it was basically a new way to architect machines. The new architecture was basically designed around the programming system and designed to minimize and simplify and accelerate programming. He figured how to do this and minimize it and make very simple machines that really worked like lightning. It was a marvelous idea.
He demonstrated it, showed how effective it was, and people got all interested in building machines on that architecture. They were proposing it to Bob -- this was right after he got to running the division -- that that should be the foundation for our future line of machines. So he asked me to lead a small group to look at the feasibility of doing this -- the hardware architecture feasibility. I was a hardware architect too as well as the software guy. So I did. We went to Hursley and we talked to a whole bunch of folks. We went back and met with Bob and said, “Technically, there’s no reason you couldn’t do this in terms of building a system or building a family of systems.” It looked like a marvelous way to build compatible systems and a range of stuff.
So Bob launched. This was after we were doing the 370, which was a modest patch-up on the 360. We were already doing that. Our future system on down the road -- five, six, eight years down -- would be the FS system. It would be a brand new system. We would no longer stick with the 360 line. Whether that was a good idea or not, he didn’t ask me. The question was: Could we build one? I wasn’t real sure it was a good idea, but there we go. In any event, it had a big crowd of people. Dick Case was brought in to lead a big task group to sort of lay out the FS plan, and they laid out a strategy for it and how they were going to do it and the way it would be programmed.
But essentially [it was] an emulating system. But it was to be a colossal system. However, it was not designed for small systems. Those guys really suffered. That’s why they later broke off a small systems group, and there’d been no preparation done for small machines. So basically the preparation for IBM’s future was all tied to FS. So basically, for a period of about two to three years, development was essentially out of gas in terms of preparing for the future. They were working on FS. That’s where all the money went.
Booch: Give me the years again?
Humphrey: 1970, 1971, 1972.
Booch: This was just right around the time that DEC was releasing their PDP-11. How much was that on your radar?
Humphrey: Oh, it was killing us. That’s what the small systems guys were worried about.
Booch: Got it.
Humphrey: I believe the FS was aimed at that. It wasn’t aimed below it. The DEC machines weren’t really low end. They were small, but they weren’t the System 3 kind of stuff. So the really small business stuff got essentially killed by this. So FS got started, and I wasn’t much involved then. I just had to run the Endicott lab. So I wasn’t involved in the FS stuff, but I was asking questions, as were others. At one point, they got it to the point where they were able to make some estimates of performance. What they discovered was performance on COBOL would be slower than the 360 by quite a bit. Nobody reacted to that, except Marketing blew up. So they escalated up. After about two to three years of a lot of work, they killed FS. But I thought it was an interesting thing, because it was a whole new system, hardware and software. I think it might have been a disaster if they’d done it. We basically lost an awful lot of time in preparing for a competitive world, and I think that badly hurt IBM for the next 10 or 15 years.
Booch: So what I’m hearing you say is that this was a foreshadowing of its stumbling in the mainframe market. Was this the beginnings of that?
Humphrey: Yes, absolutely. I think that is true.