This interview was provided courtesy of the Computer History Museum.
The Model 91 Announcement
Humphrey: My first day on the
job [as IBM Director of Programming], I went in to my office in
"Okay." So I had Dick Bevier, a great guy and sort of a number two on
this, and he was helping me, running all the other labs and this sort of thing.
So he set that up for me. I had a secretary in
Booch: Oh my gosh.
Humphrey: I couldn't believe it. And so I said, "Well, who approved this thing?" They said, "Beats me." So I called the Marketing VP, who I knew, a fellow named Jack Rogers, marvelous guy, who was in charge of all of the product marketing announcements like this. And I said, "Jack, I didn't know you had a programming group." He said, "What do you mean?" I said, "Well, Model 91 programming, we're not doing that. Are you doing it?" He said, "We certainly are not." I said, "Why are you making this announcement?" He said, "I don't know." I said, "You better withdraw it now." He said, "What do you mean? I can't." I said, "You better 'cause I'm not doing it." I said, "It's not in our plans. I've got nothing. I don't have anybody working on it there's nothing going on." I said, "This is a totaling fictitious announcement. Kill it." So he did. He sent out a wire withdrawing the announcement.
Booch: That must be unprecedented, wow.
Humphrey: Well, I tell you, it
was unprecedented, but it caused an explosion. The next day, they called a
meeting with Frank Cary in
Booch: Yeah, I remember one of the
installations for the 91 was at the
Humphrey: Yeah, that's what it was, and CDC was the big competitor, and we had to have a commercially announced product in order to bid. And that's the only way they could do it; otherwise it would be a custom bid and they couldn't get it out there and all that kind of stuff. If it was a commercial announced product they could just bid it off the shelf, but if not they would have to use special government pricing. So that's why this announcement had been hopped up by somebody in marketing and headquarters. The engineering headquarters division president had actually approved it. His name was John Haanstra.
So I stepped into the middle of this explosion, and I didn't even know it. But it was a crazy announcement -- I wasn't going to tolerate it. My job was on the line, folks. I was nailed by this thing. I wasn't going to stand for that. So this is before I'd gone through the stuff with Frank Cary on dates or anything. So in any event this meeting with Frank Cary, they were just about to start the meeting, and they were going to crucify me. I mean, they had to have this thing out for this enormously important bid. In walked a fellow named Joe Brown. He was a Model 91 Engineering Manager. He said, "Before you start the meeting I'd like to tell you something." So he explained what they called a cracked-stripe problem. It turned out over the weekend the Model 91 had developed a cracked-stripe problem, and they'd been working on over the previous couple of weeks, and they'd gotten some clues on the problem. It turned out the circuits didn't work. The chips, the way they were built, developed cracks, and they had a chemistry problem, they had all kinds of stuff, and he said, "Frankly the schedule is in serious doubt. We've got no idea how long it will take to fix it." I was saved.
Booch: Saved by physics.
Humphrey: It's better to be lucky than wise at times, but I was absolutely stark staring lucky. So all of a sudden this was cooled off, and thank God I'd withdrawn that announcement. If I hadn't withdrawn the announcement, we would've been committed to deliver the 91s on a date that wasn't anywhere near feasible. Turned out CDC got into terrible trouble, so IBM ended winning the business anyway, but it was extraordinary. So as I say, luck is something you want to have and sometimes you're down on it. But I've concluded when you count on luck you don't get it. So in any event that was that. So we went ahead we did the planning and we delivered all of it on schedule and everything worked well. That was a big change.
So things moved on awhile, and I ran programming, and we were delivering on schedule, and we were heroes and everything else and then they had another re-organization. And they restructured everything and the development division was now given to Bob Evans. And Bob Evans became my boss. This is the guy I had faced down in two different really serious crises, and all of a sudden he was my boss. I'd had battles with him over timesharing and I'd had battles with him over the FAA thing. It was the timesharing when I fought with him. I hadn't had any problems with the 360 stuff. But in any event, so they decided they were going to re-organize now, and I'd been running Programming for three or four years. This is like '69, late '69.
The Development Reorganization
Humphrey: But during the course of this whole thing I'd been promoted to VP of Technical Development, where I had programming and systems architecture, and all of that. The 370 line was coming along, and it was all before Bob Evans came and they re-organized. 360 was being delivered at a great rate, and so they were now looking at the next step which was 370. And so Bob Evans came in as my boss, and he was now looking at how to re-organize the development division. Don Gavis, who was now reporting to me running the 360 programming, came to me, and said he and his team had concluded they needed to put virtual memory in the 370, and explained it. So Bob Evans had just come in and we went to Bob and explained why we needed virtual memory in the 370. And we went through the whole thing and he bought it. He was a decisive guy, he was a brilliant guy and decisive guy. He had a problem that a lot of really brilliant people have in that he tended to view his opinions of what ought to be as fact. And so he had these facts that he was convinced of that turned out not to be supportable.
Let me finish one piece here, and that is: Bob Evans decided to re-organize IBM, which he did. But he wanted to re-organize it with each engineering department having its own programming group. So he's going to break up this big programming thing. Part of the problem had been that I had not allowed any programming product to be announced without my approval. And for me to approve it I had to have a plan on my desk. Well, it turned out that every product -- hardware or software -- had software in it. No one could announce anything -- even a feature -- without my approval, and that really frosted everybody. The power structure had completely changed. All of a sudden Humphrey was over every announcement; you couldn't make an announcement without his approval and I put some real tough constraints on it.
And of course we
weren't missing dates anymore. So this had the whole community up in arms, and
they had to fix that, and the way they fixed it of course was to break
programming up and give every engineering manager. Evans had a bunch of us in a big meeting of lab
directors, and I was there. How do you organize programming? How do you
organize the business? And my position was you've got to organize it around
programming systems. He didn't buy that. Their conclusion was we're going to
organize it around hardware systems. They had large systems, intermediate
systems and small systems, and they had system names and the System 38 and the
system this and that. I mean, so they had all this stuff and everybody got
their own programming. So that's what they did. And so all of a sudden I didn't
have a job. And so Bob decided I was going to run the laboratories in
Now the 370 had to have virtual memory, and that was what we were going to do, and we hadn't designed it yet, and the hardware guys had to go do that, and so we'll come back to that, and I took over the Endicott lab next.