An Interview with Hans Bruijnes

Hans Bruijnes


HB = Hans Bruijnes
GAM = George Michael



GAM: Today is April 14, 1994, and I'm going to be interviewing one of the earliest guys who came into the Computer Department at the Laboratory. His name is Hans Bruijnes. Hans, why don't you start by just telling us when you got to the Lab, and what your first assignment was, and that sort of stuff?

HB: Let me go to my notes. I got started in June 1953. So that was kind of early in the game, but basically the commitments for the UNIVAC had already been made. I never got into that game at all. And as you already said, I was very capable, after six months of practice, with a Marchant, the desk calculator. And we did something that was classified.

GAM: Well, I can put words to that. We were trying to calculate, for one cycle and number-by-number, all the results that were generated by a UNIVAC code, to check it out, and make sure that it was going to give the right answers.

HB: Oh�that's not particularly what I remember doing. We were trying to generate equations that fit certain expressions�equations of state and stuff like that.

GAM: Ah, yes.

HB: So the first couple of months we spent on that. I don't really know how long I worked at that.

GAM: Who hired you? Do you remember that?

HB: No, that's the one thing I can't remember. I went to work�I think Glenn Culler was the guy I reported to in the beginning. I had meetings with Glenn later on in life. As a matter of fact, his son, David, worked for us twenty-five years later.

GAM: Now he's a tenured professor at Berkeley.

HB: A tenured professor at Berkeley�so time does go by.

GAM: Yes, time does go by.

HB: You said Leona Schloss, Sid's future wife, was there early on, and Kay Purdum, and myself. And I hang out with Bud Wirsching, and we never talked about it. We meet at all the art galleries lately.

GAM: Well, I remember he was one of these persons who was trying to make marks on a card so that it would go through and be punched�electrographic marking I guess they called it.

HB: Okay, I don't remember that.

GAM: It never did work quite right.

HB: I don't remember that. Maybe that was later in the game. Anyhow, that went on for a while, and then, it says here, from '53 to '57 I was classified as a programmer. I don't know about that. I think it was, according to my records here, the summer in '54 I was working on the IBM 701, and I remember I was working for Mel Clark. He was a physicist that I was working with. We were working on a series of design programs.

I was working on STEGG, and Jim Dimmick is one of the guys I worked with. Jim Dimmick was working for IBM at the time. He now lives right near me, and I run into him occasionally. When the machine was delivered, John Griffith ran around measuring the temperature and trying to keep track of the humidity in the machine room, because climate control in machine rooms wasn't up to par yet back in those days. So he was very busy with that. And John, I guess, went off to IBM later on and spent many years in IBM. I don't know what happened afterward.

GAM: I think he retired from IBM; he stayed there all along.

HB: Yes, he probably did. I met him a number of times in later years, when we went out to accept machines. I had to get started and learn how to program, because I'd been working on this desk calculator, which is a little bit incompatible with an IBM 701. I had to learn how to punch holes in binary cards, and good stuff like that.

On the first day that I was to start on the project, Tad (Tadashi) Kishi was the guy who showed me how to make a flowchart. He rolled out this big roll of wrapping paper, you know, the 3-foot-wide brown stuff. And he said, "Now, this is how you make a flowchart." So that was my first instruction on how to program.

GAM: Great.

HB: I worked at that for quite a few years. One of the guys there was Bob Price. He sat face-to-face with me. He, later on of course, became CEO of Control Data Corporation (CDC).

GAM: Yes, Convair.

HB: Yes, Convair. He later moved to CDC. I ran into him back in the 3600 days, when he was up at CDC and had a management job.

I was still commuting in those days from Palo Alto, and so was Ozzie Palos. I've been trying to find Ozzie Palos. He's the first guy I commuted with on a daily basis, and he went to the Bank of America, I thought.

GAM: To work on the ERMA Project, yes.

HB: And then I thought he was at SRI (Stanford Research Institute), but I haven't been able to find him. I can never find anybody who knows where he went. Gil Stockwell was part of the commute pool. And Don Braff was the one who I commuted with the longest. We still communicate occasionally.

GAM: Great!

HB: Let's see, who else was there? Oh, Harold Hanerfeld, who went over to Stanford, and ended up working on the Linear Accelerator Project or something like that. I ran into him many years later, when he was a customer of ours in MFE (Magnetic Fusion Energy Program) and managed their resources. So I keep running into some of the same guys over and over again. Okay, I think that went on for a couple of years, and then in 1954 the 704 appeared.

As a result of your calling to interview me, I opened an envelope they gave me when I left the Lab, and I found out there was all sorts of exchanges trying to keep me from getting drafted into the military. And Mel Clark wrote letters saying that if they had to start over again, the whole project would be set back six months to a year. So, that was the first time I knew that they had done that.

That was in 1954, and then I guess by '56 they decided that the STEGG program wasn't going to go anywhere. Basically, I became involved with SNEGG, and I think Mel also took over SNEGG for a while until he left the Lab. I worked very closely there with Richard Green, who was a junior physicist.

GAM: Yes, I remember him. That's when I worked with him, too.

HB: Yes, and Chuck Leith, of course, who is the father of the first design simulation programs that were developed at the Lab. Chuck and I have dealt with each other many times since then, and we went on a number of business trips.

GAM: He's a great person.

HB: Yes, I enjoy him very much. It was a great experience working with him. Later on, Richard McLean worked on the FLAME Project, the weather forecasting code, with Chuck. It was in the 3600 days. We helped shoot one of the first movies made. I remember that we were looking at this movie with great pride and joy, and that we'd finally gotten the movie running. Then, in the movie, we saw this peculiar storm over Siberia. And it turned out after looking at that for some time that there was an error in the input to the code and it was a heat sink!

GAM: But what a neat way to find bugs, hmm? Just exactly right.

HB: Yes, that's a neat way to find a bug. Back in those early days, I found out one of the important things about programming. I tried to explain to Mel Clark what I was doing, what changes I was making. He'd give me physics changes and I'd try to explain the code changes to him and how they went into the program. I remember one day he came in with a little business card and handed it to me. And the business card said, "Essentially, I don't know what the hell you're talking about." And I think maybe I've remembered that message all my life. Programming has to be simpler, so that people who are basically involved in the science can be involved in the program. I think he brought that across.

In those days it was still absolute coding in binary. We were just getting to the point where we were beginning to use assemblers. Well, at first we used absolute assemblers. And later on we got into symbolic assemblers, where you could move things around a little bit instead of having to rewrite the whole thing in absolute terms.

GAM: Well, 1956 saw the arrival of the first primitive FORTRANs. I remember Bob Hughes went back and worked with an IBM group.

HB: Bob Hughes worked on the first FORTRAN compiler.

GAM: Yes, with Roy Nutt and John Backus.

HB: And I don't know quite where it all sat because we had Dorothy Monk and Kent Ellsworth. They were working on a compiler, trying to write a compiler. And I think Sam Mendicino, who basically spent many years with me and was certainly one of my great supporters, worked in that group initially for a short time.

GAM: Well, I'm going to be at great pains to try and put that story together, with Sam Mendicino, George Sutherland, Dave Storch, and you. Yes, that would be an interesting story if we can get the details pulled together.

HB: Well, anyhow, I guess that kept me going for a while. You asked about documentation. I don't know, I tried looking for documentation on these things and it's pretty hard to find. I have about a thousand viewgraphs lying around, many of them duplicates of each other. And I look at some of the viewgraphs and say, "What the hell does this mean?" Also, I found out that even a lot of the publications that we've put out, writeups about which you'd say, hey, this is useful information�we forgot to put the year on it! That kind of leaves you up the creek without a paddle.

GAM: Well, documentation was never a strong point at the Lab.

HB: It's even in brochures we put out at MFE. We put out fancy color brochures and I would look on the first page to see when it was printed, and no, we forgot to add the date in there.

Oh, there's one thing about the 701 that I remember�I think you asked me about funnies. After we'd been running the machine for about a year or so�it seemed to me quite a long time�occasionally the memory would just go kabonk, and break. It had a cathode ray tube memory, and we thought well, that was just the way this memory worked. There were problems with it, and we would do all sorts of things.

But finally one day it dawned on us that the memory the cathode ray case was facing a door that was facing south. And any time anybody opened the double door, the sunlight would hit the cathode ray tubes and reset the memory to all ones. So after that, we put a curtain in front of the memory and had no more problems. That was kind of fun. Ed Schoonover was with me at that time, and Don Braff and Doug Brainard.

GAM: Well, when you say "with me", were you already�had you been anointed then?

HB: Well, anointed in a certain sense. I think I was the chief programmer on the SNEGG code.

GAM: Ah, so.

HB: Don Braff was the junior physicist who would interface with the physicists, Doug Brainard was a programmer, and I guess Ed also was programming. I don't remember what we were all doing, but I think we were the first ones who had graphical output every morning for every production run. There was a big plotter over in the UNIVAC area. They had one of these big 3-by-3-foot things, and we put about six graphs out every morning for every production run that we made.

GAM: We had a hell of a time making that thing work.

HB: Yes, and I think that was the first one. As I say, we made it standard practice to have the graphical representation of the output in the morning. And then Don would be interfacing with the users, because we were running quite a few problems every night.

That was back in the good old days when we had no time sharing, and we had no batch processor. We had nothing, basically. You signed up on the schedule chart for how much time you needed the next day. And pretty soon SNEGG got to be a pretty important code, so we had an assigned half hour between 12:00 and 12:30 p.m. every day�I think it was the lunch hour. At lunch hour they were a little bit more generous with handing out machine time. And we'd be setting up programs for the night.

I still have a little candelabra at home that the guys made out of some brass tubing with four little birthday candles in it. That was when Liberace was a great piano player, and they gave it to me because they said I had superb capability of playing with the keys on the 709 console, entering instructions through the keyboard on the machine. We did it on the 701 and the 704.

GAM: Where did the MONITOR [1] show up?

HB: Well, that was quite a bit later. As I remember it, Clarence Badger, somewhere in the '57�'60 time frame, was the supervisor of the IBM equipment. And I theoretically reported to him, although I guess I was still a user then, but Clarence and I were getting involved and he was writing an assembler. I don't know whether it was the SAP or the NYAP-1. I'm not sure anymore. He was writing, playing with, and working on the symbolic assembler. I guess it must have been SAP that he was writing. We basically started to rely on it very, very heavily. Also, maybe even in the 701 days, we were putting together a tape that had mathematical libraries on it, like the square root, exponentials, and sine, cosine.

GAM: Yes, so you could pull them off when you needed them.

HB: So you could pull them off instead of having to write them individually. When I started, you had to write each one individually. After that, we had a library from which you could pull them in. So, we were beginning to gradually form little parts of what later became a system. To progress through this thing more or less in chronological order, I think the first thing that really took me off sideways was in 1957 when Sputnik went up.

GAM: Ah, yes.

HB: Joe Brady, Nevin Sherman, and Edna Vienop had been working on the THEMIS code, which was a double-precision planetary code for studying planetary orbits, and was an outside project more or less. There seemed to be a need to get a code that would do basically the same thing but would run a lot faster, which meant we needed to produce a single-precision code. So I sat down�and I think I was still fully involved with SNEGG�and in a week or two wrote a single-precision version of that code. Our involvement with that basically when on through 1958. I was looking at some correspondence files I still have�we were still sending out copies in '62 to various military installations where they were involved with the NASA agency.

GAM: Well, I talked to some people from Washington in that period, and it seems that neither NASA nor the Air Force had any capability to track a satellite.

HB: No.

GAM: They essentially had to come to you guys. I think that's pretty neat, and telling.

HB: Yes. We came out with a prediction as to where the Sputnik would crash, and when it would crash. We were pretty close. We were only half an Earth revolution off, or something like that.

GAM: A mere bagatelle.

HB: Which was pretty close back in those days. We were involved with the China Lake Naval ordnance. And they were interested in that particular effort because they were planning to launch a satellite from a plane�fly it up and then launch it at flight altitude from a plane with rockets. They were looking at a multistage thing. We got involved with George Sutherland, who was working down there, and we gradually added staging capability. Originally it was just a satellite orbit, but then we added staging capabilities so he could hang his motors on there, and thrust. He got into solar radiation pressures on the orbit. We spent quite a bit of time working together in that area, so that went on for a number of years. As a result of that, George, I think, liked what he saw and what he heard from Livermore, and so he came to Livermore eventually.

GAM: And he's a great person.

HB: So, that's how George Sutherland came on board, basically, through that contact.

GAM: I didn't know that.

HB: Let's see, what else was I doing then?

GAM: So, was there any official recognition of the fact that you guys were able to express the orbit of the Sputnik and nobody else was? Did the Air Force say anything, or give you a medal, or something like that?

HB: No, we didn't get a medal from the Air Force. We were interested in becoming involved with NASA and maybe adding that to the mission of the Laboratory. Anyhow, it was decided by, let me say, upper management at the Laboratory that it wasn't a worthwhile effort, and so basically we passed the code out to other installations and abandoned that project�sometime in 1960.

By this time I think I was getting pretty well involved in�according to the guys around me, it seems to me we got started a little bit earlier on this�but I went to a SHARE [2] meeting�I think that was the name of those meetings. It was an IBM users meeting. It was an annual meeting where IBM people met with users, and this one was in Seattle, Washington. And there I heard that Convair or one of the aircraft companies was working with IBM trying to develop a batch processing capability. I became interested and asked to have the cards decks sent to us. And the codes came out, and we started trying to run them. At first we had only one or two customers, but pretty soon we had quite a few people who had lined up every day. We were maybe running a couple of hours a day, and it kept expanding. At that point, Ed Schoonover, Dave Storch, Sam Mendicino, and George Sutherland became the major players in that effort, as I remember it now. From then on, we went on�that was the first major system effort in that area.

Another thing that I think is kind of interesting, in retrospect, is that once we had it going, more or less, there were two things we did. One we got chastised for. The other one I think was very interesting. We couldn't keep up with getting documentation out that described the FORTRAN compiler and other things in an adequate amount of time, so we made a first attempt at a Publisher in about 1959 or '60.

GAM: That was somewhat earlier. I remember it being called the 704 monitor. And we didn't have 704s past 1959�I mean, we had 709s by 1959.

HB: Well, I've been talking to Dave Storch, and that's the thing that bothers me, too. It seems to me that we had something earlier than that with which I was involved, but I don't know. Ginny-Virginia-Smith was the lady who came on board basically after I did the original work on the Sputnik code. Ginny took over and did the lion's share of the development and the continuous effort on that.

The 3600 came along in 1963. So Dave says that he joined us in '60. I remember Dave, at one point when I was still running SNEGG, standing there knee-deep in the printer output from the radiation printer.

GAM: No, it was from the SC 5000 printer. We didn't have the radiation printer that early.

HB: The SC 5000 sat at the far end of the room. The radiation printer, as I remember, sat right next to the 7090 in the north machine room. It was sitting parallel to the wall.

GAM: Boy, it's got me. I can't remember that.

HB: And the SC 5000 I think sat in a corner room.

GAM: Well, they threw it out of the machine room eventually, because it was throwing all that dirt around. So it was the radiation printer.

HB: It think it was the radiation printer, and I think Frank Giallanza and a couple of people played around trying to put a fan folder on it. And it worked part of the time.

GAM: But that was on the SC 5000. There was a fan folder on the radiation printer that really worked.

HB: Well, maybe I'm wrong. Maybe it was the SC 5000. Well, we ought to ask Dave Storch again to see what he can remember, because he started as an operator.

GAM: Well, he's standing knee-deep in paper�that's kind of what I remember of the SC 5000.

HB: Yes, well, he was standing knee-deep in paper, and that was my first contact with Dave, more or less. It was late at night sometime. And so he joined us later on. He said that he started about late '59 or early '60, though it must have been that we started later that year or something like that. Maybe I had Ed Schoonover working with me, Don Braff, Doug Brainard, and maybe some other people. So perhaps we got this MONITOR thing going very primitively for a while, before�and I can't find any reference yet to it�any data that would suggest when we did.

GAM: We have forty cases of manuals�all kinds of manuals from all kinds of manufacturers. And there are some IBM MONITOR manuals and stuff like that in there, and they'll be dated, so we can pin that down.

HB: Well, we ought to look it up, but so far the input I've been able to come up with is that we got going on it about 1960.

GAM: All right. You had a 1604, too, before the 3600.

HB: Yes, well that comes later, now. We really got this thing going on the 709 initially, then we moved it to a 7090. We were rewriting Jim LeBlanc's program, MOKU, at times. We were supposed to be a system group now. LeBlanc kept asking us to help adapt his code to the system. We were eager for customers, so we kept rewriting LeBlanc's code to get a customer on line. And it was a code that had half a dozen overlays in it. It was a pretty complex thing for machines in those days. I remember I got a little upset, but we needed the customers, and so we were under pressure to do both. We were trying to develop the system and trying it. And basically we took it on ourselves.

We got involved with FORTRAN, and you talked about FORTRAN. Basically about that time FORTRAN became available through IBM, and began running there. Bill Mansfield, Ann Hardy, I think, and a couple of other people started working at that time on the compiler for the STRETCH.

GAM: Oh, it could be. I remember it was a FORTRAN compiler on the STRETCH.

HB: Yes, the STRETCH was in 1961. But they were trying to improve the compiler, and they were having problems with it. It was an operating system of some sort on the STRETCH, and they were trying to improve it. Well, that was part of the game, but there was a FORTRAN compiler at that point. And we had a FORTRAN compiler on the batch processor when we started to run with the batch processor.

I think it was Bob Olivier�he was one of the customers�who had trouble with multiple arrays or something like that. So then we'd call IBM in Poughkeepsie and say, hey, this or that happened. And they'd say, oh, we'll send you a fix. So they'd twix us out a patch. We'd stick the patch in, we'd run this code that blew up yesterday, but we'd lose two codes over here. And Olivier was one of the guys who would go, "Damn it!"�screaming at us. That basically laid the foundation for our future career.

At a certain point we said, you know, this is the pits. We work our butts off, and we get screamed at for the screwups from IBM. If we're going to get blamed, let's get blamed for our own work instead of somebody else's work! You had asked when the 3600 came along.

GAM: It's got to be 1962 or '63.

HB: Yes, the development of the batch was in 1960 to 1964. Bob Hughes wasn't available to do his compiler, so we had these problems with the compiler. At that point, the people who were working on the operating system were Virginia Smith, George Sutherland, Dave Storch, and Sam Mendicino. Joan Dickinson was the secretary, and she helped us with writing the manuals. So we were basically doing a Publisher at that point. And we had the thing called the MONI Monitor; it had a little black skull hanging on our door. Ed Schoonover had left to be in charge of operations.

Before we go on to the next development, one thing that we kept doing from then on, basically, was we started collecting data within the operating system so that we could trace the cause when it would blow up. So, we had data collection in it to see what was going on, and the frequency of certain events. Some of the things that we put in there are how many source decks the guys submitted and how many FORTRAN source areas were in it, how many loads we did and how many load errors were in it, how many executions were submitted and how many executions took place, and things like that.

After a while we found out that certain people had very high failure rates and certain people had very low failure rates. That's when we made the error: We decided to publish the list! Who was the top of the list in terms of minimum failures? It turned out to be a gal with a junior college education who worked for two different physicists and who submitted the dozens of FORTRAN programs every day. And she'd have maybe one error out of ten. But there were some Ph.D. physicists who had an error for every deck they submitted! I think that's when I got called in and was told, "You can't do that!" So, we had to terminate that.

But we did continue to collect, not that kind of personalized data, but system data throughout our entire career to try to understand how the machine was behaving, how the system was behaving, where hangups and delays were, and how frequent they were. And it was amazing that sometimes even late in our career that a very large percentage of the machine time was used by people calling up the clock trying to find out what the time was. They spent more time calling the clock routine than it did in executing the physics�presumably they were trying to time how fast loops ran.

Anyhow, that was the beginning of the Publisher. So in 1960-61 I was still thinking that maybe we should start legal action to get our fair square deal, because that must have been the first Publisher that was put together.

Another thing we did at that point, and we came back to it later on, is to be able to put a phrase in, go through all of our documents that we had on line�it was probably half a dozen at that point�and scan for paragraphs; this is because in the Publisher we had to define things as logical blocks that couldn't be split over two pages. It had to be in one page. So, we'd scan the logical blocks to see which blocks contained that phrase, and so instead of your having to read all this garbage, it would just print the blocks for you that contained that phrase.

Later on, toward the end of my career, we decided that maybe documents altogether were wasteful. Basically you want to have concepts, well-defined concepts. And you do searches by looking for terms for the concept that you're trying to get clarified. And now, if you have to update data, and you change the concept, you don't have to go look at fifty different documents to find out every place there was a reference to the concept. You just fix the one definition of the concept, and that's it. You don't have to update all these damn documents. So, we got to this point basically toward the end of my career, but we were already playing with this concept back in the '60s.

GAM: In the early '60s, yes.

HB: All right. So, here we were using the 7094s and 7090s, and we kept using basically the same system. Then, in January 1963, the 1604 came on board. That was basically to allow us to transition somehow into the 3600. And the 3600, to be truthful about it, was an attempt to transition into the 6600, to basically start moving people away from the IBMs into the Control Data environment as we got ready for the 6600. There was some compatibility between the 1604 and the 3600. We stole I/O routines out of the 1604 libraries and used them on the 3600 later on.

Around the same time�in January of '63, I think�I went up to look at the 3600 and ran into Bob Price at Control Data, in Minneapolis. We went to look at the machine, and, although there were a couple of boxes sitting around, the console wasn't designed yet. There was nothing available. The 3600 machine from CDC never became a machine until it was installed in a trailer complex at Livermore. And we found out there was no system. They were talking about having an assembler.

This may be an important sideline�while the 7030 was trying to come up, George Sutherland said, "Maybe we could cross-assemble, or cross-compile, from the 7094." So we played around and we did some cross compiling. It was the first time actually we tried the game of cross compiling, and we thought it was reasonably successful. It wasn't pursued, because there was a compiler there, and there was an individual group playing with the compiler. But we were satisfied that there was some merit in that concept.

Sid [3] had said, "What can you guys do for the 3600?" So we had now looked at it and we said, "Oh, my God, there's nothing here." George had to go see his dad, so Sam, Dave, and I flew back together. I think we might have seen a little too much of the red wine or whatever it was they were serving�in those days you didn't have to buy them by the drink, and there were no limits. About the time we got to Denver, I think, we decided we were going to write a compiler. We were going to take the operating system off the 7094, we were going to move it to the 3600, and we were going to write the whole thing in FORTRAN. So, that was a step that was a little bit beyond normal.

We came back to Livermore, and that's the process that we pursued, to basically initially cross-compile, assuming that there would be an assembler there. We also cross-compiled codes from the 7094 for the 3600. Leo von Gottfried was the first one. He was running on the 1604. There was another cute little quirk in the 1604: It had a card reader that had a habit�as you stuck the card deck in and it came out the other side, the cards would float.

GAM: Ah, they were shuffled.

HB: Yes, they went through the card reader in order and they came out shuffled. One night when it happened to Leo for the nth time, he got a little upset. He took a pair of scissors that was lying there and he slammed it through his listings. The operator came racing out of the machine room. He said, "I think he's gone berserk!" These machines, by the way�the 1604 and the 3600�were in trailer complexes.

GAM: Yes, I remember.

HB: As a matter of fact, one of the trailers, I think the last one, has just been removed within the past year or two. It was across from the hangar later on. I don't know where the other one went. There were two trailers, and one was sitting there next to the big hangar. For a long time it was used by the Accounting Department; they had computers in it for a long time afterwards. But it was moved away from�what's the main building now?

GAM: Well, we have Building 111 now. Where 111 was, the trailers were right behind there.

HB: Yes, it's where the main weapons groups are now, the big building there. Is that Building 111? I always forget which one is what. That's where they were located.

Anyhow, we cross-compiled for the machine. As I say, we saw it for the first time in January, and I think it arrived on July 6. We were able to cross-compile and get Leo going on the 3600. And we gradually started rewriting the parts of the batch system, and the compiler in FORTRAN, and cross-compiled them across. Marilyn Richards came on board at that point, and she was assigned the job of writing a loader in FORTRAN. I remember I was at Fermi or somewhere making a presentation, and the head of the lab there said that it was impossible, you couldn't write a loader in FORTRAN to do binary work. The next morning he came back and apologized. He said, "Well, if you make a few changes to FORTRAN, you might be able to do that." We had put binary capability, or bit manipulation capability, into the FORTRAN compiler so we could do bit manipulation.

Anyhow, we got started. After about a month or so, we found out that the assembler wasn't doing very well, and we were having a problem with it. I think they called it the COMPASS, or maybe I have the wrong word�it was the assembler that CDC was supposed to put together for us.

So I called Dave Storch in, and I said, "Dave, can you write us�can we get an assembler?" And he said, "Sure, I'll write you an assembler." So, in ninety days Dave wrote us an assembler-written in FORTRAN again. We called it the Ultimate Assembler. Then we put a couple of different data tables in it so the assembler could assemble for the 3600, and for the PDP-6, where we had the Photostore. Before the PDP-6, we didn't have a compiler, and we needed one. So we cross-compiled for the PDP-6. We also did a little bit of work cross compiling for the 7094, but that wasn't interesting anymore. Then we modified the cross-compiler, and then cross-compiled for the 6600, eventually. As I say, I think it took Dave about sixty to ninety days to put the thing together. So we were really pretty panicky back in those days. It was about, I'd say, January 1963 that we had a complete batch system with compiler, loader, and the whole shebang running on the 3600.

At that point the second 3600 came on board�I forget when. We got a second one, and we did play an interesting game. That was around November 1963. Then, at the end of the year, the second one came in and was put in an adjacent trailer. We connected the memories. We actually ran a few problems using both memories from one processor, and discussed the possibility of dual processors working together. But we decided that was too fancy, we didn't need it, and gave up on it after playing with it. But it did run with one memory.

GAM: Well, they still haven't figured it out, so don't worry about it.

HB: I think we made a reasonable decision to defer it. Anyhow, that was the 3600 kibosh, I think. People like Jack Klingert and Frank Giallanza worked on that, and a whole bunch of people.

GAM: Going back a little bit, did you have anything to do with the LARC?

HB: No, the LARC we didn't touch on�it just went. There were a couple of interesting things, because we were still pushing the Publisher at that point, and it was getting pretty fancy.

There was one other thing that I happened to talk to George Powles about yesterday. And as a matter of fact I got kind of panicky to try and remember, and I called Virginia Smith. I hadn't talked to her in twenty years, but I called her last night and spent about an hour on the phone with her. And all three of us remembered the same thing: We got into voice, basically.

GAM: I remember that.

HB: There was Jack Oliver, Roger Fulton, and David Mapes. They were working on digitizing and things like that. We decided that we could make an analog trace of a voice, and record it on film, and then digitize it. So we digitized it, and we had a 3-bit register on the 3600. We tied into that. So we brought the digitized data back in from tape, then we ran it through the 3 bits, basically getting an analog output, and fed it into a little amplifier.

At an open house, we had one of these tape recorders sitting there. That was the only amplifier we could lay our hands on. We had that plugged in, and we had a sign above it that said, "See�no tapes!" For the open house we would run some program, I think it was FLAME for Chuck Leith. And we'd wiggle a bunch of tapes in the machine, and then it would say, "Oh, this is the 3600 speaking. Let us explain what we're doing here, what I'm up to." And then we'd go through some little speech. Then it would end up and say, "Well, I've got to get back to work," and boom. Everybody walked in, just listened to it, and walked out. Nobody appreciated that we were producing the voice from the computer!

GAM: Revolutionary, yes.

HB: Now that I think about it, that was probably one of the earliest digital tape recordings that we'd ever seen. It is now a common thing.

GAM: That's right.

HB: So we were only thirty years ahead of schedule.

GAM: Well, it's just sort of lovely testimony to what you can do if you are just left alone and know what you're doing.

HB: Well, there was another thing, and I think that you were involved in that. I think it was on the 704 or the 709 that we had a cathode ray tube on the machine for making film output. And we tried generating a matrix, 100 x 100 or something, of bits, and photographing it, and then scanning it back in. There was a scanner on the tube. And we thought that maybe we could record information on film...

GAM: Memory.

HB: ...as a high-density memory device, and then scan it back.

GAM: That's right, we did that.

HB: I think you were involved in that. And I think that didn't go anywhere else, except that it lead to the Photostore, eventually.

GAM: Yes, it lead to the Photostore.

HB: So we knew the Photostore was a great idea. We'd already thought of it ten years before.

GAM: They did it much better than we did. I must give them credit.

HB: Yes, but we were there a little earlier.

GAM: Yes, that's true. That Photostore was a remarkable machine.

HB: Yes, those were some of the things that I remember that we did on the 3600. Oh, there was one other thing that we worked on. I'd lost the code. Ginny wrote a code, I think, that looked at written language, and phonetically analyzed it. The intent was, and we played with this for quite a while, that we were going to have vowels and phonemes digitized and stored, and then we would be able to read text...

GAM: String them together, yes.

HB: ...and read text, giving computer voice capability. As I say, we worked on that for quite a while. Ginny remembered doing that, too. We also had music. That was a project we worked on for quite a while, and then it just got out of hand. We had other things that had to be done, and these were all futuristic ideas, so we set it aside. And the amazing thing now, I guess, is that the Macintosh is coming out with that kind of capability.

GAM: Yes. Well, remember, you were fiddling around on the PDP-1, too. You were trying to fix up a tape that would tell the operators, by voice, "Hang this, remove that," that sort of stuff. We did all that work on the PDP-1 just to get it ready.

HB: Yes. And then we came back finally and we implemented that at MFE. [4] All the instructions to the operators were spoken.

GAM: I heard that, yes.

HB: And the idea there was if you had one operator on at night, they could walk around, and if something was wrong, they would hear the thing announce that the network is down. Then they could decide if what they were working on was more important than getting the network back up. And that was very useful to us there, so we were way ahead on that count. Anyhow, that pretty well covers what we did on the 3600. We had a lot of ideas and we got a long way down the line.

GAM: I think so.

HB: I think you were asking me what my positions were at the Lab. I was a programmer until 1957. Then I was a supervisor of the Batch System Development from 1960 to '64. Then in 1964 I became Division Leader of the System Development Group. And that was about the time that they asked me to see what we could do with the 6600s and improve on their performance. I think the first thing basically that got us involved is that we had a batch system running, and we had happy users�or relatively happy users�on the IBM 7094s. We had a large crowd of users that worked exclusively on the 3600, that were running through batch. So those people wanted to make use of the 6600. So we initially installed it within the existing great omnipotent benefactor.

GAM: Generous, omnipotent benefactor�GOB!

HB: Well, I remembered great, omnipotent benefactor�anyhow, GOB. So, we installed it on GOB and ran on GOB for a while. You know, you have problems here and there with any system that's being developed, and you have to make modifications. GOB was written in machine language code.

GAM: It wasn't even machine language. It was a special set of macros that Bob Abbott had put together�just a series of macros.

HB: Oh, take that off! I was going to just say it was written in machine language.

GAM: All right.

HB: Anyhow, it was hard to update. So, as we had to make changes, as the workload on the machine became heavier, and as we tried to remove bottlenecks and get more good work going, and correct things, every time we made a change we'd have a hard time getting it implemented. Finally, we decided that we might as well upgrade it also to FORTRAN.

GAM: FROST [5] is born.

HB: FROST is born. Judy Ford and Pierre Dubois were the ones who spent most of the time on that. By this time, now, Sam Mendicino was basically just focused totally on the FORTRAN compiler issues, and CHATRAN.

GAM: You mean Judge Judy Ford?

HB: Honorable Judy Ford.

GAM: Great!

HB: As a matter of fact, Judy worked for Bob Abbott for a while, too, before she joined our effort, and Pierre Dubois. I think they put it together.

GAM: Oh, yes. You're right.

HB: Judy worked for Bob Abbott for a while on this stuff; so anyhow, we got that done. I forget now how long it took us. But that gave us FROST, like you said. At that point, I think, we had come up with a system by about 1965. It was totally written in a higher-level language, was time-shared, and had batch capability. It was well ahead of anything else in the industry anywhere.

GAM: That's quite true.

HB: No question about it. In fact, it was written in a higher-level language, but then, ten, fifteen, or twenty years later, people started talking about UNIX as being a higher-level language system. My God, by that time it was old stuff. So that was that part of the system.

Then I guess about that time, by 1968�it's kind of hard to go into details�I was promoted again to Assistant Department Head for Planning. Bill Mansfield basically took over my function in charge of the System Development Group, and George Sutherland became the one who was responsible individually to get the 7600 on line at that point. He was the coordinator for the 7600. But basically he had the individual groups already�Sam with the compiler, Dave was concentrating a lot on I/O structures, trying to improve disk usage performance. George was the coordinator. I can't remember�there must have been people who were working on libraries and other things, but I don't remember who they were or what they did.

Ed Schoonover by this time was in charge of all operations. I have to get hold of Ed one of these days. I don't remember any more how he got to being in charge of operations and when that happened, and how we got out of the SNEGG environment.

GAM: Well, I'll ask him at some point.

HB: Yes, we'll have to have coffee anyhow, one of these days. I'm getting a little interested in this now, I think.

GAM: Well, that's what happens. Now, along about this time, also, Sid Fernbach was working the political scene in Washington so as to bring MFE here.

HB: Yes, you're right. For the next two or three years, basically, from 1968 on, I was doing evaluation of the ASC machine from TI (Texas Instruments), IBM 360-195, and the STAR program from CDC. We were looking at benchmarks, and developed the Livermore kernels that later became very popular, but basically developed it at that time. I was collecting some things. Frank McMahon became a mainstay within the compiler group for quite a period of time. He invented Stacklib, which later allowed us to put a vector declaration into the compiler. Frank was very much involved in these activities. Then later on he carried it on. Frank has been basically my source of information as to what the possibilities were and what he thought of machines for a long time since then. I can give you his current estimate on the massively parallel if you want it. It's not good.

GAM: I know it.

HB: He was very much instrumental in helping me out at that point. There were a couple of other things that we did. I think it was about the time that we got the first RJETs out, and I know I was focusing on those quite a bit. There were about a dozen RJETs (remote job entry terminals) throughout the Laboratory.

So we went back then for a while, and I played with the business of digital voice. I thought maybe again we could play the voice game. And instead of trying to read random information like we had played at before, we actually dictated all the FORTRAN expressions, the alphabet, and the numbers into a system. So now when it looked at a FORTRAN listing, it knew sine, cosine, plus, do, and other expressions. It would recognize those and use those words. So, basically it would read FORTRAN to you. That was reasonably effective.

There's a fellow, and I forget who it was. He was in charge of the Theoretical Group, he was blind, and he had a Braille Teletype.

GAM: Jim Slagle.

HB: Yes, Jim Slagle. And that started me thinking about it. I said, "Well, gee, a Braille�that's kind of inconvenient. If you could have it read to you it would be easier." Again, that project worked reasonably, except when you got the memory dumps. You can't sit there and listen for an hour to a memory dump where it says, "One-zero-zero-zero-one-one-one-two." And that needed some deep thought on how to handle that. Again, we ran out of time. You have these little ideas, and you play with them for a while, and then you give up.

We did a lot of work to bring graphics to the RJETs. Dieter Fuss tells me also that at that time he, Ervie Ferris, and I were playing with a Tektronix display. I think it was a memory scope of some sort�a visual. We had that down in Sherwood, and we were sending graphs out to it as an addition to an RJET. I think we installed the first RJET out there, and I think that we were playing with the idea to add that to it. That's how Dieter and I met, I think, working on that project.

Again, Marilyn Richards also was working on the 6600�the G machine�and by that time it had been put out where the LARC had been. We were playing games with changing the table sizes at night, so that at night the table sizes would shrink as the number of customers disappeared. That way, there would be more memory left for large customer production runs in the daytime, when interactive service started up and usually the individual demand for memory wasn't as large but we had more customers. Then we'd start increasing tables to keep track of the various customers. That was the only system we ever developed basically that had variable table sizes to allow large programs to run at night and more programs in the daytime. It never got into the system that I know of later on.

GAM: That was a good idea.

HB: Yes, it was a good idea. But now you can buy so much memory, it doesn't make any difference anymore, anyhow.

GAM: I guess I'd agree with that, yes. I'm not sure that's a neat solution, but it works.

HB: No, but for a lot of things that we wouldn't have considered neat solutions, now you just throw a little bit more money in the pot and buy it.

GAM: If you only have a hammer, everything looks like a nail.

HB: Yes. So we played with those things there for a while. And we came back, by the way, to that digitized voice idea. I tried that one more time in about 1985. There was a company called VOX, I think, in Fremont, and they sold boards that allowed you to read information at random. We actually had our electronic mail. You could call in to a telephone number, and type in your pin number. It would open your user number, open your electronic mail, and read your electronic mail items to you.

GAM: I used that, but was tiring.

HB: Well, the reading and speaking weren't all that bad. But the problem was that it would usually go through the address and the telephone number and the head listings, and that was boring as hell. It would read digits individually, not as numbers, and in time there would need to be more coding added to it.

GAM: Well, it's pretty good, though.

HB: Peter Pearson worked on that later on. Anyhow, we got through with the RJET activity, and then Sid came to me and said, "All right, they're interested in a fusion energy computer center, and it would be kind of nice if we could get that." That was in 1973, and Dieter Fuss had been working on the German translator, another one of his ideas. We had a German translator going, and Dieter and I were collaborating on that. So we knew each other reasonably well.

So, Sid asked me in my capacity to put a proposal together, and Dieter had been working on the Sherwood Project, so he was familiar with that. So we worked as a team and made a proposal. I think it was in September 1973, and by October we were selected as the MFE Computer Center, or the CTRCC. But Livermore was selected as being the fusion energy support center. Then later on in the following year John Killeen came on board, and we got started. So that's how the thing got started initially.

GAM: Well, your approach there as far as system development is concerned was different from what it was at Livermore, or inside. I mean, it had, first of all, telecommunications to worry about to a large extent.

HB: Yes, we had a lot more network you mean?

GAM: And you had a different kind of serving computers�PDP-10s, PDP-11s.

HB: Well, because there was a fairly strong local capability that was tied in with the network to our central capability�so that kept evolving. And that, in looking back at it now, would take a lot more time to really research exactly what all the various stages were. I get a little confused about all the old terms.

GAM: I don't know about that. I remember a lot of it. I remember you telling me that one of the things that had been planned was that people would compile on their local PDP-10, and then send the stuff over to the fast machines. They found out that they could go a hundred times faster on the fast machine�so the hell with the PDP-10.

HB: We knew that from the beginning.

GAM: Well, yes, but the people didn't. I think that's great.

HB: Initially we used the 6600 from about the middle of 1974�July of '74.

GAM: You used the 6600 down at MFE?

HB: Yes. When we used it, it was out in the building where the LARC used to be. That's where the 6600, the G machine�serial 1�it was sitting there. And we were using it to support unclassified work. Then, when this came along, we started using it. And we used it from I think about July of '74 until the fall of '75, because the procurement for the 7600 took a long time. First we had to put the documents together for the procurement, and go through the whole procurement cycle, and then take delivery on the 7600. So, the 7600 came along in October 1975. It became available through Dialup, but the delivery actually was in May 1975.

So, in May or June of '75 the machine became available, and then we had to work our butts off for the rest of that year. And as I remember, it wasn't until basically '76 that we were on line fully with the network, and the PDP-6s and all that "Mickey Mouse" stuff. But we had to put the network in, as you said, and develop it, and run around to all the remote sites. We had to build a new, independent archival storage base that we didn't have. So we had to install a 6400 to run the archival storage base. We had a 3850�the CDC machine. So there were a lot of odds and ends to resolve in trying to start an operation from scratch. That kept us pretty busy.

I think we brought in the 7600 running LTSS, the same system that they were running inside. We might have had to make some modifications for I/O drivers and things like that, but the archival storage was the thing that basically took a lot of our time. And the network took a lot of our time. Jim Leighton, Cliff Cordova, who's since died, and Gary Hendricks, were some of the guys who put in a lot of time in the beginning to get us off the ground on that one. That was a really hard job. Then again, Paul Lund and Dave Storch were on board at that point with me, and other people came in. It's kind of hard to remember everyone who was there.

GAM: Doug Kent.

HB: Doug Kent came in, yes, and Lee Tennant. Yes, I've got a list a mile long. As a matter of fact, I came across a telephone list, and I said, "What the hell time was this one from?" I've got a call into Jean Shuler because there was a Joe Choy on the list.

GAM: Yes, I remember Joe.

HB: Joe Choy was working with her on trying to put the Mass Store thing together. And then he moved.

GAM: To NCAR in Boulder.

HB: Yes, NCAR. And he was on this telephone list. And Louisa Louie was on it. She was a gal who joined us in the beginning, with Steve Lewis. They were among the first, but also people like Art Scott, Barry Howard, and Peter Pearson. There's a whole bunch of people on this list. But I was surprised, because I know Louisa Louie left fairly early in the game. It all must have been in that 7600 time frame, so I'm still trying to run down when all those people came on board. There are about thirty people on that telephone list. It was kind of a surprise that I had a telephone list and that Joe Choy was on it.

GAM: Well, they were very important people. They really did some important work. I think one of the unusual things down at MFE was the successful installation of the satellite antenna.

HB: Clement Luk was another one. Yes, the antenna came a little bit later. First, Larry Berdahl came on board. Although I've had some differences of opinion with Larry Berdahl later in life. With quite a few of these people, you know, you may have differences of opinion, but they certainly made very serious contributions.

After we got the 7600 up and running, we had to go on to the next level. The inside was focused on the STAR at that point. And, because of inside information, I knew that it was a bad proposition, because the scalar speed on the machine was too slow. Everything you gained in the vector area, you lost in the scalar area. So I was already aware of that. There were some people who didn't believe that, and I won't discuss who.

GAM: Yes, I understand.

HB: But Dieter and I were pretty well convinced that it wasn't the way to go. And there were some reservations at that time about dealing with Cray, a new startup. But they won.

GAM: I think that was a very good decision.

HB: Well, I didn't think it was a bad decision. Anyhow, we went with the Cray, but that now meant that we had to again make a major jump from one equipment type to another equipment type. That meant that the compiler that Cray would furnish us wouldn't do the things we needed to do. It didn't have the bit and bite manipulation, and all that sort of stuff.

We redid our compiler to be able to cross-compile everything for the 6600. That kept us busy for a while. And, as I say, guys like Larry Berdahl were instrumental in that. Then there was a whole bunch of people working on stuff like that�Clement Luk working on libraries and I/O structures. So we were very busy getting ready for the new Cray. But, again, we succeeded, and the machine came through the door. We ran on the Cray system for a short time�I think it was three or four months. Then we decided that we could roll over and put our own operating system back onto it, and take off from there. Dieter might have a better memory as to how long it took to get that show on the road.

Fred Fritsch and Dennis Lawrence took over on the Publisher. And the only thing is, I remember that in 1968 I was interested in getting something done that would be more terminal-interactive. We were just moving into the terminal era, so I thought that we should be able to do things on the terminal, and modify reports. Then, it was still very much oriented toward a massive input from card decks or existing dates, with update decks and so on.

GAM: The whole documentation system was in the Photostore, and I believe that the TRIX thing�and then Gary Long did TRIX GL.

HB: Well, it might be. I wasn't familiar with...

GAM: But, I mean you could read the documentation on line on your TMDS (television monitor display system).

HB: Okay, yes. Well, I guess I was focusing on other things when that was running. I do remember that we came back to it in the MFE era. So at about 1975 or '78, we were interested in getting our documentation on line again, and it had all disappeared. Now, I don't know what decisions were made. There were big manuals written about the Publisher, and a lot of detail. And I think it was decided that it wasn't the way to go, because we had to start from scratch. We started out with a new Publisher again, and we had to recreate all the documents we'd had on line years before, rewriting from scratch. I was looking for card decks to be able to restore these files to our system.

GAM: Interesting.

HB: Fred might know what happened to it�where it went. But I have a feeling that TID didn't feel that that was the way to do business, and that's what killed it.

GAM: Well, I don't really think it mattered what TID thought, you know. It's not clear they thought very much.

HB: Somewhere towards the end of this outline there are machine notes that I put together. I think about the time I left we had a system running that was basically totally compatible with UNIX, but had its own good parts left. We still had an automatic dump of anything that blew up. We don't have that now, unless you put in special requests.

GAM: So, when did you retire?

HB: In 1989.

We had a system that was itself running in parallel. To my understanding, the current system they're using on the Cray machines still has no parallelism in itself. It offers parallelism to the customer, but the system itself is not running in parallel.

GAM: Right. That's true.

HB: But the Cray system we had�the CTSS system (Cray time sharing system)�was running in parallel, and it had a number of things you could run in parallel. And it offered parallelism. The parallelism that was being offered by Cray was very poor. It basically was meaningless.

GAM: Well, I think that CTSS survived longest down at Los Alamos National Laboratory. I believe they kept it going for a long time.

HB: But, you know, you said TID had no input, but Washington just told us, "No more. You will now go with vendor-supplied software." I think that was a terrible error, and a terrible oversight, and a terrible lack of understanding of what the realities of system technology were.

GAM: That's what you get when you have amateurs making technical decisions. That's what those guys are in Washington.

HB: I think that you're considering their capability well above where I think it is. A businessman would go back and get some input, I think.

GAM: Maybe they're just Harvard MBAs.

HB: There you go.

GAM: Yes, I think they meddle far too much in the technical stuff.

HB: Yes. I have some thoughts written here. I don't know where they came from. I don't know whether I invented them myself, but I typed them out because I found them, and I thought they were kind of cute:

"Success isn't a matter of intelligence�it's a matter of determination."

"The bright guys are in cafes discussing how the world should be; dumb guys are in the office changing the world."

"Surviving international competition takes a lot of tenacity�and lots of tequila."

GAM: Well, that sounds good. What is this thing that you're referring to all the time�this document?

HB: Oh, this is just�I told you that over the last couple of weeks, every time I ran into you I felt I really wanted to get some notes together about what had gone on.

GAM: I hope you'll leave those with me. [6] Well, I think we've had a good session. Let's stop here.



[1] MONITOR was a batch operating system supplied by IBM and extensively modified by us.

[2] SHARE was an IBM-sponsored user organization through which certain standards were developed and experiences were shared.

[3] Sid Fernbach, Head of the Computation Department.

[4] MFECC was the acronym for Magnetic Fusion Energy Computer Center; "MFE" was the common name.

[5] FROST was the acronym for FORTRAN Resident Operating System for Time Sharing.

[6] Editor's note: Hans Bruijnes' notes had less to do with the Computer Center and more to do with friendships; therefore, they are not included here.