An Interview with Nevin Sherman

Nevin Sherman


NS = Nevin Sherman
GAM = George Michael


NOTE ADDED: One of the standard rules followed in producing these interviews was that the person being interviewed was requested to review the typescript, to ensure that it said, as accurately as possible, what he or she intended to say. Despite several attempts, we have been unsuccessful in getting with this person to complete a review. There may be a variety of reasons for this to be so, but the effect is always an uncompleted interview. So, the reader is warned that this interview is not finished, and awaits possible further editing, if and when the interviewee wishes to do so.


GAM: The date today is 7/15/96, and we are going to interview Nevin Sherman, one of the first persons at the lab. I guess the best way to begin is to tell us where you came from and when.

NS: The University of California, Berkeley.

GAM: The fact is you were in the Astronomy Department there. What year was that?

NS: I can't remember the last year, it was a couple of years though.

GAM: But 1952 was one of them surely?

NS: Well that was probably the last one, or maybe even 1951.

GAM: OK, so then you came out to the Lab, do you remember when you started here?

NS: No, unfortunately, I don't remember, but it was about that time.

GAM: Late '52 or early '53?

NS: Something like that, the UNIVAC was here.

GAM: Oh, the UNIVAC, which means it was after April, 1953.

NS: Originally, I probably would have been on the list of people who went back for training but because my papers weren't processed, I didn't get to go. Bob Jastrow was supposed to take care of my papers, but something happened that he didn't get it done.

GAM: So you determined the person who forgot to process your papers was Bob Jastrow. So you didn't go back on the UNIVAC trip then?

NS: No.

GAM: OK, so what happened?

NS: You mean the gory details?

GAM: Yes.

NS: Well, I had made an agreement with him that I would be gone for a month and a half, I would come back with a P-clearance and I would be ready to get on the payroll. And I would need the money because I would come back broke. He assured me there would be no problem—he would take care of everything. So I left and when I came back, I was broke, I kept my part of the bargain, and he looked at me and he said, "I'm sorry, I forgot," which was a surprising response because it was so honest. But, it was also very aggravating and so he offered to start the process again, which would take another month and a half at least. In the meantime, I would have to get another part time job, which I did. Then finally when they did call me in I said, "No"—I didn't want to go to work there. They were pretty peeved—they are probably still peeved about it. I forgot about it and, sometime later, I reapplied and an administrator at Berkeley, I believe by the name of Evans, an older man, looked at me and said "Why should we hire you, you already turned us down?" So I explained to him exactly what had happened and why I turned them down and he left it up in the air as to whether I'd get hired or not. A couple of weeks later, I guess, I got a notice saying, "OK". They'd forgiven me for my misdeeds.

I went through the process but, of course, I'd missed all that good stuff going back to the UNIVAC class, getting prepped, etc. When I got to the Lab which, as everybody knows, was a pretty shabby place at that time, no air conditioning, and no amenities of any sort. They had really hard water that had been in the pipes for five years. I was in the cooler for a year, the better part of a year, before I was cleared. I don't know why it was, I finally had to go up to San Francisco and be interviewed by the FBI. That's where I found out what was bothering them—of course, in those days, anything bothered them. The trouble was, you didn't know about it unless you went up there and talked to them, you might never know why they turned you down, if they turned you down. The outcome of it was that the FBI was satisfied that my Dad wasn't subversive and that I wasn't.

It turns out one of the big things against me was that I had been a member of a co-op in Berkeley. It was just a living co-op to save money, but these guys were thought to be Communist sympathizers, I mean, in the worst scenario. They had some wild people up there to some extent, but they weren't Communists, of course. The FBI didn't know them personally and they get some information and then try to construct a picture from experiences you know yourself, and they might come up with the wrong conclusion. Anyway, that's what held me up.

I also found out that my Dad, who was a professor back east, had signed some things along the way during his career. A thing like a petition or something like that, and it turned out it was something that was on one of the lists of the House Un-American Activities Commission.

GAM: What was he a professor of?

NS: He ended up being a Professor of Spanish. He was a language professor.

GAM: The romance languages?

NS: Originally, yes, but then he gradually narrowed it down. He was pretty neat, I sort of appreciated him, but I appreciate him a lot more now that I don't have him around. He was a remarkable guy, honest as hell. Even in the small schools there was a lot of politics, and I didn't appreciate that. He hated it, but he had to deal with it every day.

GAM: Where did you go to college before Berkeley?

NS: I went to Pennsylvania.

GAM: The University of Pennsylvania?

NS: No, Franklin and Marshall College.

GAM: Oh yes? That's a great school.

NS: Oh? It depends on your point of view, I guess. Actually, it was good. I wanted to go there, but I didn't realize until afterwards that it was pretty darn good. It was a small school, and it was a good liberal arts college. I think that, by hindsight, it was better than I realized. Although, at the time, there wasn't much in astronomy and science. There was a real watershed around World War II. It's unbelievable. Prior to the war, when I was back at that school, there were probably three people in the whole city that you could call scientists and they were professors. The city might have been 60,000, the size of Livermore now; there was nobody else, and you couldn't go off and talk about anything interesting. The average person didn't know anything about it. Of course, I didn't either, none of us did. In those days, calculus was a sophomore subject in college and I thought it was pretty darn difficult. In fact, it threw me for a loop, it's just the way they did things, I don't know why they did Calculus that way, but they had a lot better methodology in ordinary things that we neglect now, like English. They didn't talk about Language skills in English Literature. It was a different era and you can't get those things back and I don't know if you should. But, the science part was pitiful by comparison of what we have today. It's incredible.

GAM: Well, it's interesting how you found your way out to Berkeley then. Which is the epitome of science stuff.

NS: Well, I suppose, but I just applied and they accepted me and that's the only reason I came out here. I knew I didn't want to work in a factory. I'd worked in a factory in the summertime and I saw those poor devils; I'm not looking down on them, they had to get their jollies every place else because they couldn't get them on the job. It was a grinding job and they didn't pay the minimum wage they talk about now. All the prices were comparable but, anyway, I just realized I didn't want to do that, I didn't know what I wanted to do.

I don't know how you grew up, but I didn't have touch with many people in science or anything else. I just didn't have the slightest idea about what it was, but as I said, I think World War II just changed everything so much. Then, of course, when Sputnik came along, people like Christine, who were in school just prior to and following Sputnik were lucky. There was a real acceleration into warp space or whatever they call it. It's wonderful, I look at the books available now and they are unbelievable. I used to look at the very popular books available in the library and then some tomes, there was nothing in between, there were no good astronomy texts, they were very, very mathematical or very much for the specialists or very popular. Now they've got this marvelous selection. I mean in Astronomy, which was certainly a backwater then, they didn't have those things and Astrophysics was basically almost non-existent anyway.

GAM: You couldn't even spell it.

NS: Well, nobody knew and then, of course, for a long time it wasn't even accepted because it was so speculative. They were lucky if they got the right sign to the exponential, they didn't have to worry if they were even a magnitude off, just going in the right direction. Anyway, it's a wonderful place, these days. But, there's so much other garbage if you were a youngster. Now there's such a terrific pile of manure and everything else with it, it's like a needle in a haystack. But it's all there, where before it wasn't there at all, there was much less to sift through, but it wasn't there for you once you sifted through it. In one sense, I'd love to be able to start over. I wouldn't want to throw away what I have—it was too dang hard getting it. But I enjoyed getting it. But the things weren't accessible then, it's amazing now.

GAM: You had a good experience under Leland Cunningham, he was a great person.

NS: He was an unusual person.

GAM: That's better to say than great.

NS: Well, he was both, you probably heard all the stories from Joe Brady.

GAM: I also knew Cunningham because he used to come out here to use the computers.

NS: He didn't come out here for long, unfortunately.

GAM: Well, he got the 701, and they put a core memory on it, and he was happy as heck.

NS: Well, he grew up when I grew up and he cut his teeth on plug boards. It was all IBM machines and, if you could do that kind of junk, it is really tedious. I remember when I was at Berkeley. That's what Joe had to do for his thesis. You had to make a difference table and you'd wire a board that was just differencing. So you'd put all these cards through this darn hopper and once you got the tons of them through you got your first differences. Then you'd take those first difference cards which were one less than the number. You'd close and lock the door so wind wouldn't blow, you hope; and you'd put them down on the triangular array, one last difference out here and you'd have to pick up the right cards and put them through the integration route. I remember thinking: Oh God, if that is what you have to do, forget it, I'm out of here. But computing didn't exist basically, in those days except for that. Joe's thesis was a remarkable work but one of the things that will always stand out about it was the way the HOD had to be done. At the time, it was the results you were interesting in, and they were very worthwhile. But the remarkable thing is actually when you think about what they physically had to do to do this, it's actually amusing. They had to wire every board to do every little operation and process these things. It was incredible. It was really an amazing job.

GAM: So, while you missed going back to UNIVAC school, you got the UNIVAC C-10 code and you read through it and learned how to program on the UNIVAC?

NS: The fellow I was trying to get you to remember came out and gave me the code, so I started studying it. I spent months on it and it was sort of tedious and dry and I didn't understand. There was nothing about programming that I could understand, because programming didn't exist. Anyway at the site, we didn't know what it was. Well, it turned out that the only thing in it that was related to programming was in the last page and a half or two pages of the darned thing. I should have turned there immediately, but I didn't know this. So I learned a lot of interesting things about the UNIVAC. How to read the numbers in excess 3, and all kinds of details about the mercury delay line memories, which was fascinating. Then there was another guy in the cooler who never did a darn thing. He came in much later and some guy came in and spent a day with him and brought him up to speed whereas it took me months to get there. Boy was I mad, his name was Ben, the fellow in the cooler with me. It was fascinating, very interesting.

When they got to the LARC, they published a programming manual. Of course, by then, programming had developed. But when they got the programming manual they didn't have the people who designed the machine make the programming manual. They had an outside group who were really ignorant, about the way a beginning programmer would be, starting without knowing too much and trying to understand it and present it in that way. When you get an expert to write something about his work, it's going to be great for another expert but it's not going to be a good general purpose manual. That's unfortunately what the UNIVAC manual was. I found out later that the engineers who had designed it had written the manual. Of course they talked about what they knew, but the things that puzzled me, they didn't even worry about.

GAM: I had a tough time with it too.

NS: Once I got to the page and a half, or two pages that gave an example of a loop—I don't know if you remember or not, but you had to program your own B Boxes. You had to pick up a thing and modify the address and then test it to see if it was greater or less than something and then back in and then transfer—actually, it was a wonderful machine in some ways. It had a lot of drawbacks, but it was a wonderful machine. If it had an assembler, it would have been nice, but you had to take every memory location, it was a number, and you had to keep up, oh it was hell trying to remember what you had used that for and if you over-laid things and so on. Then, with only a thousand word memory, that thing was a bear for me. But you got two instructions per word and it was often numeric which is very nice. But keeping track of stuff was tough and putting in patches and then trying to remove them later—they needed a program that detects it. I remember one program, I don't remember the name of it, but somebody programmed it and then left the area. Someone else was assigned to try and fix it up. They removed one little patch and the thing never ran again. They never could get it to run. Those were the days—an assembler would have been a wonderful thing. Anything that was symbolic, that you could address so you'd at least know what you were talking about when you gave an address.

GAM: Actually, I made a table that equated names to addresses and I used to refer to that when I was coding.

NS: Yes, that's what everybody tried to do, but pretty soon your scratching, it's just like rewriting the tenth draft over the first one and you have to make sure you don't wipe out something you were trying to hold there.

When we got on those two-dimensional problems, Doug Gardner and I called it TONY. We had fifty-nine overlays. And there'd be some error committed or something committed in one overlay that would show up ten overlays later. You'd get in an overflow all of a sudden and have some digression, you'd have to trace back though that or this. I'm telling you.

GAM: For the benefit of the record here, TONY was a two-dimensional hydrodynamics problem. Did it have heat flow in it?

NS: Yes, it had heat flow. Actually, it was fun developing the equations, but it was very tough to write the program.

GAM: I remember TONY was a big problem. Probably the biggest one that ever ran on the UNIVAC.

NS: Well, there was even a three-dimensional one, but I don't know what ever happened to it. It was called HERCULES.

GAM: I never knew anything about that one.

NS: There's a funny episode here. It turns out that TONY was done for B-Division. In particular, it was John Foster's group and John Foster was eager to run the problem. They were working with some guy who set it up originally. He wasn't here long, and I didn't appreciate him at the time but, by hindsight, that guy was probably one of the best programmers of that era. Because he worked out the original scheme for this thing and it was so simple. I don't know if I would have ever thought of it or not; it was really neat. Doug Gardener and I had to take over when he left and we had a lot of trouble getting it to work.

If you recall, debugging was a real problem too because you'd have to wait. Something else would be on, like Bob Abbott and his group would have a problem running and you couldn't interrupt it. You'd have to wait for a break. I'd wait around fifteen hours a day to get a few minutes at four o'clock in the morning or something like that.

GAM: We were learning.

NS: We were learning, but the thing is, we didn't have any turn-around scheduled, and that was amazing. So many things were amazing about modern computing, but turn-around is one of the things you take for granted and there wasn't any to speak of and it was tough.

But getting back to my thing about Foster, do you remember the creamery in town?

GAM: Yes

NS: Well, we used to go there for lunch. In fact, Teller even showed up there a couple of times. I'd go in there and Foster, I don't know why, but he'd be out thumbing a ride by one of the gates, I knew who he was so I'd pick him up. He always seemed so friendly as though he knew me and I thought, "well, that's nice, he's a friendly guy." It wasn't too many years later, in fact this was in the early 50's and I didn't find this out until about 1970, he had written us, both Doug and me, a real commendation for the work on TONY. Of course it was put in our files, and in those days you weren't able to see your files. He said that he didn't expect us to finish it and it was remarkable that we got it working finally and that we should get a raise. We didn't see any of that, of course, and we didn't hear that we'd done a good job. It would have been nice to have someone say "Hey, you did a good job", but nobody did that. I only found out when I got laid off and I heard there was such a file. I asked to see it, and they didn't want me to see it, but they let me. I said, "well, I'll Xerox some of it" and they said "No, you have to look at it here and you can take notes." Well, you know you can't do that but I read through it. I never saw the letter. I wish they had let us know they could have deleted the part about the raise.

GAM: Yes, well, I think that's nice that he did it all.

NS: Yes it was, but what I think is not nice is the intermediaries didn't let us know, that's all I'm saying. I didn't expect him to do it and I'm glad he did and it was nice. It was really neat.

GAM: So you worked on TONY with Doug Gardener? Colonel Douglas Gardener, great, great person.

NS: Yes he is, he's still around I'm sure.

GAM: Yes, I see him now and then when he comes out here. I'd like to interview him too, but I haven't caught him yet.

Anyway, after TONY, what did you do?

NS: It's hard for me to remember exactly what we worked on.

GAM: Do you remember when we had that big open area? It wasn't the cooler, it was just a bullpen like and everybody sat in these little glass cubicles. Someone came in one night and put crepe paper down your windows and it looked like bars.

NS: Yes, I remember that, and a sign that said "please don't feed the animals" or something like that.

I don't know if you remember this or not, but there used to be a big open air area with chairs and tables and a lot of us were in the same room. There weren't even partitions. I remember John Hudson shouting at the top of his lungs and saying the half word so he wouldn't have to go to an overlay. He'd be so excited he'd make you feel like you had to take the afternoon off or something like that. He was really nice.

GAM: John's dead now. He retired and moved to Oregon.

NS: You're kidding, no I didn't know that. I'd heard he'd gone to Seattle, but you are probably right, I heard it second or third hand.

GAM: He was a gentle, nice man.

NS: He sure was. He had a big hand in STEP.

GAM: John's contribution was the Monte Carlo techniques and stuff like that.

NS: Dick Butley, too, and others like him. Butley once told me, he was resigned to it, but he said, "we don't get any credit for anything we do. We can't publish, obviously, because it's classified." So, basically, what they did was a black box to the people outside. So, in some ways they didn't get some professional credit for what they did, just to augment their career a little.

GAM: I think Dick did a great number of things, finding the uneven distribution in the random number generator that we had. It was very interesting and very illuminating about how to generate random numbers, a strange, strange, art.

NS: You had to really think about what it meant to be random. I don't think I thought about at all. I don't think I do now.

Ever since I looked at the random log, I had some idea of what it is. That's what got me started thinking about it. What a random log is and then how you did well. That doesn't tell you much about generating random numbers, just the use of random numbers.

GAM: I remember once, I think it was Von Neuman and Teller, picked up the New York telephone book and grabbed a number out of it and that became the seed for one of first random number generators. Someone said that wasn't really safe.

NS: No, no. Well, as long as the sequence was long, the repetition longer than what you need, it was probably OK. But if you started to use up a reasonably large fraction of that kind of stuff you are in serious trouble.

GAM: Around 1956, or so, we went and specified what turned out to be the UNIVAC Livermore Automatic Research Calculator (LARC). We also used to call it the "Leith Atmospheric Research Calculator".

NS: It was also called the "Leith Altar". Bernie said that.

GAM: Right, those two guys got the most use of it for awhile. It was a very interesting machine, but it took too long to get it.

NS: Oh, it was my favorite machine.

GAM: It rewarded the person who took the trouble to put in his code.

NS: But it wasn't hard to do, that's the thing. You take an IBM 650 machine. To be efficient on one of them, you had to be careful where your program was on the drum You had all kinds of trouble with the word boundary. The LARC was just simple. It really was simple to use efficiently. It hadn't the drawbacks with memory because there were still in tubes.

GAM: Maybe it was core memory then?

NS: Didn't the memory involve tubes?

GAM: Those were the drivers, but the actual memory device was ferrite cores.

NS: The memory was cores, sure. The drivers were the tubes, and that's where all the problems arose. If it weren't for that the thing would have run without any errors.

GAM: Well, the rest of the machine was the first transistor machine we had.

NS: Yes, I remember something from one of the engineers when they came here to install it. He said that this was supposed to be, I think they had hoped it would be, the first fully transistorized machine. But the transistorized stuff couldn't drive the memory.

There are a number of things I remember. Of course, it was two years late, but one of the things I remember they told me was every place they didn't meet the specs, they exceeded them. Whereas, when they got to the IBM STRETCH, they didn't, and they had to cut the price by 20 or 30 percent.

GAM: Actually, it was about 48 percent.

NS: Oh, was it that much? I didn't know it was that much. Of course they had a lot more of them, and it was a much later machine.

GAM: You know, Nevin, it's not so much a lot more, well maybe, they built two LARC's. One went to David Taylor Model Basin (DTMB). There was one at the factory and one at Livermore. There were 11 STRETCH's.

NS: Oh, yes, I know. They were slated to have more, maybe.

GAM: Well, the STRETCH fell flat on its face.

NS: I didn't know it was that bad, but I knew it was substantial.

GAM: It was very complicated to program also.

NS: That's what I've been told, I never tried it.

GAM: To put things into perspective. Joe Brady had been working on his THEMIS code that was supposed to do orbit calculations...

NS: More than that. You could call it that, that's a good enough definition. It had a unique feature that I was so happy about. In those days, we didn't know much about orbit harmonics. Nobody knew much about that stuff. We are talking now about less than a year before the launch of Sputnik. 1957, I remember, because we got a paper from back East about this that I glanced over. But the thing was how do you allow for the obliqueness of the Earth? Well, there is actually a technique called Ivory's Method which enables you to take a SAUD, a spheroid, and instead of getting it in that series, you get an exact principle, an exact result. So, what we did, we took that, we took one big Spheroid then subtract one slightly smaller, so now we had a shell. Then you'd take that one and then another. Anyway, that was the idea and I thought it was pretty neat. So we'd get the array broke down. The only problem was, nobody had thought the Earth as much a shell, you really didn't know what numbers to put in. And there was nothing to fit it to because satellites hadn't been invented yet. It was pretty good, in that sense I thought it was pretty neat. I'm just guessing it might have been the first code, although there's been some big firsts in that era.

GAM: Well, I think it's close.

NS: Well, it was probably one of the first, if not the first, codes that accounted for the perturbations of the oblique Earth. The trouble was our hand was forced, we didn't get a lot of time or effort on it until there was a satellite. Then, all of a sudden, hey, I want results. This is what Sid kept bugging Joe about. I didn't hear it, but he'd say to Joe, "Well, why can't you produce this orbit?" Joe would point out that if you don't have initial conditions you couldn't do anything. Obviously, that's part of it. Then Fernbach would say, "How does JPL get these things out?"

This was Pickering, it wasn't William Pickering, Thomas Pickering or something like that, who was the head of JPL. He was Australian, I remember, and he was the first head of JPL. Joe said to Fernbach, "Well, why don't you call and get some initial conditions?" He called them, and the guy claimed they didn't have any. Joe says, "You mean he was lying?" and I said, "I don't know." Well, we never did get this kind of thing.

There's one other thing you have to have, and this takes a lot of effort. It's nice to have an orbit about the center value there, let's say, that would be reasonable, but you really want to relate it to observation. So, you have to have a fairly good mechanism for linking up the observer, and this is not totally trivial, you're not observing it from an inertial frame, you're observing it from an observer's frame called a relative coordinate. It takes a lot of doing, you can't do it in just one half day. What had happened is we didn't have the authority, or the time, or investment in time, to do this initially. So, all of a sudden, when we need it they wanted it the day before yesterday, and you can't do that. That was a serious problem. Well, I suppose everyone got caught flat footed by Sputnik as far as that is concerned.

GAM: Well, NASA certainly did. They had no way of calculating that orbit.

NS: Do you remember October 4, 1957?

GAM: No.

NS: Well, I do, I remember exactly. I used to live at the UNIVAC basically, and Jack Oliver, who was our technician, came in for the four o'clock shift. He had heard the news, and it was really news. He came in and told me, and you have to remember that there were hardly any rumors that they were going to do this. You think, well, this is going to leak out, but it really didn't. And, all of a sudden, here's this thing beeping and it is really incredible. Well, the outcome of it all is he knew I was interested in this type of thing, and he was trying to let me know. It took about an hour before he convinced me that he wasn't putting me on. Then, I was like everybody else, I was just flabbergasted, and it was so wonderful.

GAM: I remember your reaction.

NS: In those days, I was commuting to Berkeley a lot, to and fro and I'd think about this. It would take me an hour and a half to get to Berkeley and I'd be thinking about this thing. God, I'd be going from UCRL to Berkeley and that thing was circling the Earth in 90 minutes. Oh, it was incredible.

One night, Joe and I were trying to work on this program, THEMIS, to try to get it to do lifetimes. It's not misusing it, but you don't need all the refinements of the purturbations to get lifetimes, that's pretty crude stuff, but that's all we could do because we didn't have the initial coordinates and all that.

I remember we took a night off in February and we drove North of Livermore. In those days it was quiet and dark. You'd get about four miles away and it was pretty dark. We knew about where to see this thing in the Northwest and it was the first time I'd ever seen a satellite, obviously. You don't think about looking for this thing in advance, in detail, and the shadow of the Earth is up there, which of course you can't see. So this thing is actually pretty high above the horizon, but inside the shadow, and all of a sudden it comes out and it's going five miles a second. In a matter of two seconds, it's way out in the total darkness and it was pulsating because it was rotating and bright. It was so incredible. And then, when you realize in the case of the second one, there wasn't a human life in it, but there was an animal. The second one was bigger and brighter. That was an incredible experience, I'll never forget those days.

GAM: I remember, especially, you had stars in your eyes.

NS: Well, it knocked me out! People grow up with different things but, as a child, because of the astronomy, you start thinking about planets and space and life. The first thing is you think there's no life in it. I used to watch the moon and it's incredible. You can dream of cities on the moon and things like that. Bu,t eventually, you start to realize, "Hey, this is very unlikely". It is a vacuum, of course, you could still have chosen life, there are lots of things you can imagine, but it gets more and more remote as you get older. Then, all of a sudden, you start to think, "Well, as much as I'd like to, I probably won't see rockets, or space travel, or even the concept of the satellite in my lifetime". Then, all of a sudden, it's here, that's just amazing to me. Then you get complacent, and think you can do anything and here we are 30 years later, and while we've done a lot, we haven't done anything like we could have done. You do something, and then you slack off for 25 years because it isn't socially acceptable for one reason or another.

GAM: Well, as I heard the story, NASA came to the lab and gave us an additional 704 to do this orbit calculation because they couldn't do it. Then Hans Bruijnes and someone else worked on a single precision one? What was that all about?

NS: Well, just an adaptation of stuff that we'd been doing, but it was on a faster machine. Still but it had tapes on it. It still made errors, but they weren't as numerous. But, they were undetected ones and we put drag into it and everything else.

It's funny how things stick into your mind that have nothing to do with the main topic exactly. One of them was that we had something like one minus the cosine of Theta, and Theta is a small angle. So, you don't want to take one minus the cosine of Theta because the one is awful significant. The problem you have in astronomy and integration is maintaining significant figures and, if you aren't careful, you lose everything. Now you don't even know it because it slants everything one way or the other and changes the exponent and off you go and, if it doesn't overflow, you don't know that you've lost some significance. Well, Virginia Smith, remember her? She was a very pretty woman and very capable. But I could never get her to change. I said, "I don't want to write one minus the cosine. I don't want you to write the program that way". In those days, you could go into the subroutine with the lead term plus one. That's basically what you had to do, but I could never get them to do that. So, for the very small angles you couldn't calculate when you were representing observations. If your prediction was close to what was observed, your absolute care made the difference. So you could make a correction. Well you couldn't do that, not with our technique, unless you went to double precision. That was easier than figuring out an alternative. Basically, that's where they did some lifetimes predictions.

Another thing I remember about Jastrow, just in passing, the only time I ever ran into him again. I never talked to him again after that initial time. It is 1969, or so, and by that time he is at NASA. There was a guy by the name of Joel Brenner from SRI. He was a Mathematician. Somehow, I don't know how, we met him. He came over here and he and I and McGoven and Roger Fulton. Remember?

GAM: I remember Roger.

NS: Roger did the programming, I think, for the 704, but I'm not sure what machine it was exactly. But, it was about 1970 so I could probably find out. He suggested doing some calculations. The second U.S. satellite was this little grapefruit that Nikita Khrushchev was always calling "that little piece of junk". Well, that's the one that actually detected and effectively measured the second and a third harmonic in the Earth's gravitational field. It was actually a terrific experiment, but there wasn't any communication with NASA and I didn't know how they did it. It turns out that the calculations on which this was based were done under the supervision of a man named John O'Keefe. I don't know if you remember that name, but he's a very bright and remarkable guy. A real idea man. Anyway, they did these calculations and we didn't have the slightest idea how they did them and they probably couldn't tell. So, we decided to try an alternative, "Could you change the way this eccentricity operated without the third harmonic". Without going into any detail, we found out we could do it. The only problem, it was not both a necessary and sufficient condition. You could do it that way, but that doesn't mean it was done that way. That's what I wanted to say. I wanted to say that this wasn't the best way to do it. Joel wrote the paper, basically, and they went alphabetically. He said, "Well, you can sign it or not. If you don't want credit, don't sign it". Well, I wanted to get some credit, but I said, "I think this is too strong, you don't have the right to say this". You should say that, in lieu of the fact that we don't know what they did do, if you made your observations in this manner, you would come up with this fit which is similar to what they had without the third harmonic. Well, he got it published and, about two weeks after it was published, he got a call from Jastrow and he was furious that we were impugning that the work that O'Keee's group had done wasn't correct. I told him, "You didn't have the right to be as strong as you were". That's what I said. Anyway, Jastrow, talk about suing us, he said, "I can put you in touch with the SRI lawyers". But, he sputtered and quit. But that was one of the things I liked about Jastrow when I first met him. He was scrupulously honest. Maybe he was just too darn dumb to make an excuse, but I don't think so. By hindsight, I was always impressed by that honesty.

GAM: What was the calculation you were trying to perform on the LARC?

NS: There were two. They were related, but one was a general N-body problem. And N was 380 and that was pretty big for those days. Of course, it was soon superceded. It was Duane Kinmann who did the programming, and I tried to derive the formulas that we needed beforehand. I don't know why we got to do it, but Spencer Manlove came and offered the chance. He was head of the LARC. He asked me if I'd like to do it and I said, "Sure I'd like to do it." He said, "I'll have to keep this quiet because I don't think Fernbach will necessarily approve." He said, "You can have one person to help you." He probably mentioned Duane, I don't know if I knew him before that.

GAM: That's who the other person was?

NS: Yes, so we got started on it, and people always want you to estimate how long it will take you and what it will do. Fortunately, I always survived by compensating errors. Parts I thought would cause me trouble didn't, but the parts I thought would be a snap often were very troublesome. But, it all balanced out because I overestimated on some and underestimated on others. We got the thing going, and I think it was pretty good, but it wasn't anything but a prototype. You have to remember, we were still using assembly language. That's one thing. The second thing was we didn't have time to finish the big question in N body problems of close approaches. When it's either small or intermediate, when it's large sometimes you can gloss over these things, the close approaches. And that's a tough call. It's not easily solved but we had a crude way of handling it at that time. It worked for us, although I wouldn't recommend it in general. I wouldn't think it would work. For the N-body problem there are 3N second-order differential equations, so you need 6N constants if you want to solve it completely. All those constants are integrals. There's only 10 known, in general, and one of them, of course, is the energy interval. Physicists love conservation and the conservation of energy is one of them. In this case, there's no closed space in energy in principle. So, we used that, we looked at the full energy. I'm not sure exactly how, but every umpteen steps, we would sum the energy and we would get it to stay constant up to eight or nine significant figures. Now, I wouldn't recommend that, it's a necessary, but not a sufficient condition. That's one thing. The second thing is, as you have more and more bodies one of them can really dance around erratically, and it won't show in the sum of 379 others, and that's the danger. But we didn't have time to think about it much. So that's what we did, and we monitored it, and if we saw it jump, we had a re-start routine that would take you back to before the time at which the trouble had occurred and cut the interval, reintegrate past the it, and restart it again. It worked enough that we got something like results for a hundred and twenty million years or something like that. It was interesting. The difference between ours and others that came later, we tried to follow objects carefully rather than saying, "Well, it doesn't matter." You had the option of saying, "Well, maybe there are little errors but they aren't significant." If you follow a trajectory accurately, if you really do that, then you automatically don't have to worry about the other thing. Otherwise, you have to wave your arms and say, "Well, this is a meaningful result even if we made these little errors here, but nobody really knows." We tried to be careful.

And the second one was a double precision integration of the solar system. In fact, it would have been something Joe Brady could have used in his thing with Planet X, if you remember Planet 10, we called it. The number of tapes limited us—we were going to produce a consistent set of the planetary coordinates. So, if you wanted a representation of Haley's Comet, or picking out any other object, the source of the planetary coordinates for perturbations would be right there on the tape, you wouldn't have to compute them. The tapes would be dated—you'd pick up what you need. If you didn't want to use all the planets, you didn't need all the tapes. Eventually, it blew up because there was no reliability you could trust. Our tape drives had been changed for IBM drives, if you remember. Eventually, I had to abandon this because while I could run it, I couldn't guarantee to produce anything usable. It's important to be consistent internally as much as it is for accuracy.

GAM: So, what was the thing that raised your suspicions that there was an error somewhere?

NS: Oh that, yes. Something was published in a thing called the Magnet, a publication for the Berkeley Rad Lab. I had found a LARC arithmetic error by differencing. If you came through the UC Berkeley Astronomy Department, you learned about differences and about high-order integration. For instance, a third derivative is not a third difference. That's just the lead term of an infinite series of differences. Well, people in hydrodynamics like Bill Noe didn't know that or, if they knew it, they forgot it. Christine, one time, almost worked for that group. She didn't go because she thought she'd be first out if anything went wrong. She thought she was being set up, not deliberately, but if she was the last one hired, she'd be the first one let go and they were having problems. They had some hydro set of equations that had a second-order derivative and a third-order derivative. For the third-order derivative, they just put the third difference. For the second-order derivative, they put the second difference, but they didn't have the second difference to the same accuracy, the same significance, the same order.

Well, she talked to me about this and I said, "Come on, you can see what they are doing". And she said yes, and she went and talked to Bill and he doesn't realize what he is doing and he says, "Oh, a correction term" as though it were some fudge factor. It was a correction term, but it wasn't his kind of correction term. They didn't seem to realize that kind of thing. All these things are infinite processes, unfortunately, and your only hope is that the derivatives would come up in terms of powers of something. You only hope these things are small enough that the high powers would knock down these coefficients to solve high order ordinary differential equations. When you come in with the partials, you have to cut everything to a minimum so you'd high powers will knock down the coefficients. It made me realize that these guys were never required to, really have to take just a relatively few differences for everything. I mean you can't use high-order approximations here you'd never get off the ground. He didn't realize that.

GAM: So, you found an inconsistency by differences?

NS: Yes, and, with the LARC, if you multiplied the member location times the contents of the register, this could give you an error of one unit in an umpteenth place. When I discovered this, I wasn't working in fixed point. I was working in floating point, but it would have been just as true in fixed point. I thought it came up one unit in the sixteenth place, but I could be wrong about that, and just exactly one. But, then, if you reverse the role—if you put whatever was in A, (the accumulator) in the memory and multiply it, then it would be right. Basically, when you multiply things you have to shift and add, shift and add, right?

GAM: Right.

NS: That's basically what you're doing, and in the case of the positive numbers you're stuffing the high-order end with zeros and I think when you multiply by negatives you're essentially stuffing them with nines, something like that. And he said something to the effect that the way the circuitry was designed it was possible for it to "forget" that it was using a negative number. Stuff it with the wrong animal, something like that. They had a heck of a job, it took them a long time, and they had to make several passes at this thing and you had to go through these thick mats of wires—incredible. What was it a bore scope or something like that they used?

GAM: Yes, but all things considered, it's a remarkable achievement. To find a logic error at that late date.

NS: You know what it reminds me of? I don't know if the analogy is very good, but do you remember the Pentium Processor thing within the last year? The big to do, and they said, "Yeah, but it won't come up very often"? It shouldn't, probably, but it still did, and that's sort of true here too to some extent, because it didn't happen very often.

The other thing, my part, after discovering I was being absolutely sure, because nobody believed me and you're never quite sure and you don't want to go off half cocked. But, then, I made up a fairly elaborate, for me it was elaborate, routine check without using multiplies. Take a random number and sop it in and pick up another one for the second half word, and do this for two double-precision words, that kind of thing. Then it would do the multiplies with adds and subtracts, and rather than any multiply circuit, and then it would do the actual multiply with the multiply unit and then compare the two. If they agreed it would keep a tally, a count and say, "OK, you got one success" and you'd keep going on and you'd have the time between successes or failures. When I'd run this with just arbitrary numbers, I mean generating a scatter graph, it would go on and on and on. It wouldn't make many errors, it wouldn't happen very often. I just happened to look out and if I hadn't looked at the difference table, wouldn't have found this out, how would I know? It's really a matter of luck, that's what I'm saying.

GAM: No, I don't think so. I think you have this nature of wanting to check stuff and this bias you get from your astronomical studies.

NS: Cunningham was the guy, he was amazing. He had sets of formulae, if you went through textbooks and studied you'd come up with some sort of formulae. He actually was very cautious and he was talking about the days of hand calculations. The most he could ever get were ten significant figures. You want to do a lot of ten figure numbers, to add six or seven figures after that at the end. He would very carefully select what formula he would use to actually generate the result. The other half, which I hated when you had to do it by hand, was trying to check it. You never did it by doing it over; you did it by using another set of less significant formulas. You could estimate what the maximum error was and that's what you went through. It was a terrible job.

GAM: But it's done, I think that's great!

NS: I didn't mean it was luck, but it was still a fluke, I just happened to be looking at differencing where this occurred. In the old days, when you hand calculated you worked from difference tables. You had to because there wasn't anything else to do. You didn't have any other memory except the paper. When you get to the machinery stuff you can't look at most of it because it spewed out this prodigious stuff. Anyway, that's the way it happened.

GAM: I think that's fine, I don't want to make light of it though. I believe the prepared mind will find this kind of stuff out.

NS: You know, one incident I just thought of, was on the UNIVAC. There was a lot of competition for a little spare time on it and Cunningham had some. I don't know what he was doing. There was an eighth satellite of Jupiter and all the outer satellites are so distant from Jupiter that the perterbations just of the Sun are immense and these things don't begin to close, it's just very rough, it won't repeat itself. There's no really sharp point of significance and you can't follow them with any analytic formulae, nobody knows how to do it. So, you had to do a numerical integration and often Jupiter is not in a position you can use. It has to be in opposition every couple of years and it disappears behind the Sun and if you don't follow this carefully it'll get far enough away from it's predicted position and you probably won't find it. So, it got lost. In fact it's really Famous, the name of our problem, it was named after the tenth satellite of Jupiter. There was supposedly a tenth satellite of Jupiter that was lost and that was its name I believe. Of course, it was a minor deity in Mythology too.

GAM: What was the name now?

NS: Famous. It was a minor deity in either Greek or Roman Mythology. The question is, when they discovered things, did they really discover something and then lose it, or did they just make a mistake. They didn't have star fields and photographs and things like that, it was terribly difficult. If you see in a field of view a hundred thousand stars, it's very tough, without our mechanisms of modern technology, to keep up with that stuff. I think that's why we picked that name.

Another thing I remember about that program was, since you were at liberty to pick units, why should you pick the radius of the Earth with some odd-ball number? Why not pick something convenient like 4000 miles? So that's what we picked. When you'd talk about Famous, Cunningham used to hate this, he'd say, "That strange unit you have." But the thing I liked about it was that, if they said 7.2, you say immediately what it was in miles if you wanted to visualize it. I thought that was sort of neat, I don't know if anybody else did. I don't think you did. It was unconventional but it didn't make any difference.

Getting back to this eighth satellite, well it had been lost. There was a guy by the name of Herb Grosh, who was a strange guy. He used to stand up on a chair at meetings, because he was a short guy. I talked to him a couple of times—he was a neat guy, a funny guy. I wouldn't want to get on his wrong side. He had done the last good piece, a very meticulous piece of work on the numerical integration of this eighth satellite. It was done in the days when it was still by hand, just in that transition period in the 50's or somewhere in there. Anyway, it was lost and somebody had to do something about it. I don't remember who proposed it, but I remember Joe and I were both interested in throwing this up on the UNIVAC. Tom Wilder apparently even said, "I'll help those boys" so they won't waste your machine time. They gave it consideration and then Fernbach told me, "It's not a UNIVAC problem". I thought it was, but I went off and there was something called the "CPC".

GAM: The Card Programmed Calculator, right next to the UNIVAC.

NS: I had to do tons and tons of work if I was going to use that, which they let me use. I had to do it on my own time and do an unbelievable amount of work just to try to get the stuff set up to do the project. Anyway, about six months later, there was an announcement from the East Coast that they had recovered the eighth satellite of Jupiter and of course that it was done on the UNIVAC. It was done by Paul Hergut, who was an ace programmer. He was sort of like Cunningham, I think. I don't know how much they were alike, but they had the same role. I'm guessing he was a little bit younger than Cunningham was so he was probably a little bit more into "modern machinery" than Cunningham was.

GAM: Cunningham was remarkable, I remember he would stand behind an operator with his notebook orchestrating all this stuff on the UNIVAC.

NS: In one sense, he was one of the worst teachers I ever had. But I learned more from that guy because I had to do it myself. One of the things I really liked about him, especially by hindsight, years later, now that he's gone, he would decide we would derive something by elementary means. That meant it was going to be a lot of work. You had to do a lot of detail and try to check it. He always committed to start something on the board and go through it and at the end we'd get some intermediate result that wasn't right. I'd go over my notes and try to figure out what it was and do it again and I'd come into his office at night and I never wanted to say, "You know I think you made a mistake", because you didn't do that with Cunningham. So I'd go into his office and I'd hem and haw and I'd say, "you know that thing we were trying to develop? Maybe it should be so and so". And he'd say, "Well, don't we always to it that way"? I wanted to say, "You made a mistake", but I was never man enough to say that.

This has been positively therapeutic for me.

GAM: Well Good. I want to thank you for spending the time to take me on this marvelous trip through your early years at the Lab.




For information about this page, contact us at: comment@computer-history.info