Newsletter of the Computer History Association of California
ISSN 1071-6351
Volume 3, Number 1, November 1995

Kip Crosby, Managing Editor

From Vacuum Tube Flip-Flops to the Singing Altair
An Interview with Lee Felsenstein (Part 1)

[Lee Felsenstein - sysadmin of the pioneering wide-area network Community Memory, contractor for Processor Technology, hard-working moderator of the Homebrew Computer Club, and designer of the Pennywhistle modem and the Osborne One - has lived a life constantly interwoven with the history and philosophy of computing. Through it all he's pursued Ivan illich's ideal of "conviviality," insisting that technology attains its highest purpose only when it becomes understandable, approachable, repairable and usable by ordinary people. This first part of a projected three-part interview takes us from 1959, and a computer club without a computer, to 1975, Homebrew, and Steve Dompier's history-making implementation of "Fool on the Hill."]


KC: Let's start with flip-flops in high school....

LF: Yes, at Central High School in Philadelphia; it was the college prep public high school - actually the third high school in the nation, founded in 1838. Until 1983 it was a boys' school but it's since gone co-ed under pressure of a court decision. But in 1959 my brother Joe was a senior and I a freshman, and he founded the Central High School Computer Club. He had been looking in books from the forties that he found in the library, with vague circuits in them of gates and flip-flops, and he wanted to build a computer - I think because he wanted to program one. Of course computers in those days had a tremendous cachet, they were UNIVACs, the "giant brains," they would occupy half a floor of a building. He would show me these books and say, "Can you build this, can you build this?" - and while I had my doubts, my relationship with him was such that it wasn't legitimate for me to say no, or so I thought; it was just my position as a younger brother, to always be tagging along. So he founded the club and installed me as the chief engineer, and we would take 6SN7 tubes - dual triodes - and mount them in sockets in cigar boxes. I had a basement workshop at that point, I was familiar with the tools, and I even had a rather crude audio oscilloscope. And we tried to make flip-flops so we could make logic elements, counters and registers and so forth. Probably they would have worked quite well, except that I was completely ignorant of so many fine points. For example, we were trying to create pulses for counting by flicking wires against Fahnestock clips on the cigar boxes - and I can't think of any better way to create a random stream of pulses than that! But that point never entered my mind and I didn't really have any mentors who would point this out. We certainly could have used help from an engineer, preferably in the computer business, but we never really considered going out and asking for it. It had only been two years since Sputnik, and the school was dominated by a competitive aspect that ruled us all. I mean, some of the boys were keeping running grade point averages up to three decimal points in the backs of their notebooks, that sort of thing. We were all supposed to go claw and scramble over each other and get into the best colleges. Actually, that kind of thing had been happening at Central High School for generations.

In any case, the Central High School Computer Club managed to produce a couple of two-bit counters that probably worked about fifty per cent of the time, using our techniques. I concluded from that whole experience that computers were not for me. I knew I had a future in electronics - I wanted to be an electronics technician at that point - and that gradually evolved into, well, I'm going to college so I might as well study electrical engineering. I had gotten into electricity and electronics at the age of 12, when I read in a chemistry book that if you wanted to be a chemist you had to learn a lot of mathematics, and I was having trouble with percentages at that time. Only much too late in the game did I learn that electrical engineering required the most math of any engineering. So much for career planning! My experience with the Central High School Computer Club made computers simply too frightening for me, from a hardware standpoint, because they had to get it right every time, every clock pulse. I had no particular interest in the software, so that turned me away from computers for a while. (My brother, incidentally, went on to become a professor of genetics, specializing in mathematics of genetics, population genetics and so forth.)


My next experience even close to computers was in 1965, when I worked with what was called the Free Student Union, which was the successor to the Free Speech Movement at Berkeley. I learned how to punch cards on an [IBM] 026 card punch. You could walk right into the card punch room and use it without any questions asked, so I punched and sorted the membership lists and we used them for local meetings that nobody came to.

At the end of 1967 I had dropped out of school, I was in a psychological depression, and in the beginning of '68 I went off to work as a junior engineer at the Special Products Division of Ampex Corporation; then at the end of 1969 I dropped out of that job, and then dropped back in when my money ran out.

Ampex is almost gone now, but it was a big company in those days and a division might have two or three thousand people - the Magnetic Tape Division as an example. The Special Products Division, which was a hundred and fifty people in a little building on the Redwood City campus, took responsibility from beginning to end for one-of-a-kind products. We might just take an item from stock and paint it purple, or we might design something from the ground up. We'd gotten into the design of a large-scale audio-visual instructional system, called a Pyramid system, controlled by a [Data General] Nova minicomputer. About three of these were built. Well, in 1970 I was put on this project - that I didn't want to be on, because of that computer controlling it - and I was rather quickly given responsibility for catching the bugs in designs that other people had prototyped, and then teaching other people about how they worked and to support the ongoing effort; it's called maintenance of product design, MPD. I started with the tone detectors, which were nice comfortable analog things; but I moved on to something I had done previously, which was design of audio circuits for high speed tape duplicators - IS oscillators at 1 megahertz, that sort of thing. In effect it was hi-fi design at 40 times frequency step-up, which was very interesting. But now I couldn't avoid getting into computers....

KC: Right.

LF: And first of all I was sent to class to learn BASIC at the Service Bureau Corporation, SBC, which was the setting for some very important realizations. The instructors at that place were young guys in three-piece suits who liked to show off how much they knew. And they would say, "You see how the system sort of hiccupped there and it's running more slowly? That's because the computer in Los Angeles has been turned off and it's been transferred over to the computer in Kansas City." This was my first experience with networked computers, and I understood that geography could be irrelevant on a computer network, providing it extended to the various places. They also showed me how files could be made public with levels of accessibility, by pre-pending asterisks to filenames; three asterisks on a filename and anybody in the system could get the file - in effect, it had been published. So I saw a hierarchy of distributed publication possible there.

KC: Now, what operating system was this?

LF: I don't know what operating system it was. I presume that SBC had won some kind of anti-trust suit with IBM and somehow gotten an IBM network out of it, probably running on [System/] 360's, but I don't know the details beyond that and you may have to look them up if anybody's interested. Almost certainly these were IBM computers, we used Selectric terminals, 2741's. All I knew was, I was learning BASIC. But I still had some important realizations there, and several of them related to media and publishing. I had spent much of my time in Berkeley working for the underground press, on the grounds that this was a certain kind of community media - media with less hierarchy. There was still a hierarchy, by definition, because that's built into any publication system, whether print or broadcast, where the information is replicated identically and sent out from a single point. In fact I had given up on the underground press because it was subject to many of the same tendencies as the mainstream press, including centralization, and the temptation to make a spectacle of itself in order to increase its circulation. That wasn't the game I was interested in playing, or assisting. I wanted to help define, and help create, media that contributed to the development of a local community. And now I was presented with networking technology that could be a foundation for communities of interest - communities that could exist in defiance of geography. We could, at least in theory, untie the knot that tied a single person to a single community. This was a vision that I wanted to elaborate, and continued to pursue.

Meanwhile, at Ampex I also learned enough machine-language programming on the Nova to do little drivers and so forth. By the end of 1970 I did the programming to demonstrate the control structure between, if I recall, 12 key pads and 12 buffer tape recorders. The system worked through a series of master recorders, each of which had 8 tracks on quarter-inch tape and could run bi-directionally at high speeds. A matrix switch with reed relays sent output to a buffer recorder at each student position, or carrel. Each carrel had earphones, a video monitor, and a 12-button key pad whose interface I had designed. You would sit down at the position and punch up the number of a program you wanted. That would set up a transfer between the track on the relevant master unit and your buffer unit. So the master would energize and do a high-speed tape duplication, 40 times speed, across to your local machine. In a minute and a half [of duplication] you'd have an hour's worth of program copied over. I think the transfer was carried out backwards, so that when the duplication finished, your buffer machine was at the beginning and would start playing, and you would hear the program material on the earphones. Buried in the program, meanwhile, were periodic bursts of 55 Hz tone which a tone detector picked out and turned into a slow serial bit string; and this would command the computer to do various things - not only tape motion commands, but also primarily transfers of pictures from a moving-head video disk recorder with a 14-inch platter to a set of buffer tracks on another video disk, so you'd get a still picture transferred from a repertoire of about 300 still pictures. That picture might present you with a question, and the tape would be stopped, until you responded on the keypad. The keypad, like the rest of the system, was interfaced to the computer, and depending upon what your answer was and the instructional program that was running, the tape would either continue with the right answer, or rewind to another track and play back an explanation of why you had been wrong. It's interesting that, by 1990, I was able to build all this functionality into a unit that attaches to your belt, works off a compact disk, has a private-eye display and headphones, and audio synthesis and speech synthesis and audio playback, and a little pointer that you hold in your hand - an isometric pointer; so that's a closing of a circle, in effect. But the machine language programming that I had done, that allowed the buttons on the keypads to command the tape recorders to do the transfer and then start playing, saved my job [at Ampex] during a reduction in force that was going on.

So I gradually became more familiar with computers, through designing interfaces. I also did a design analysis program in BASIC for the active filter components in the tone detector. Those things are easy enough to design, certainly out of cookbooks; the hard part is to build them so that the filters' performance remains acceptable while components range over their tolerances. I had been given a design with ideal values, and as part of MPD I was supposed to figure out how to build it with actual parts so that it would work every time without adjustments.

KC: In other words, it'll work this way if it's perfect, but what about when it's not perfect?

LF: That's right. Engineering is mostly about how far off the ideal you can stray and still be safe. So I constructed a test suite with a lot of nested loops that would vary each component through its possible range - that's on the inside loop - then go out and vary the next component one tick, go back in and vary the first one through its range. There was a nested loop for every component. Running this in BASIC on the Nova, I was able to demonstrate that no possible set of combinations for that design would work over its entire range. Meanwhile, of course the units that had already gone in the field - the prototypes, in effect - were giving all kinds of trouble. So they sent out an engineer who was clever enough, and simply didn't want to hear about anything else; and he sent back a cardboard extension to the board wired with potentiometers and one-shots and some other parts. He had done almost everything that I wasn't allowed to do, and I was told "Okay, just put that in the design and build it." Naturally we did wind up with adjustments, and that was an excellent lesson in the relationship between engineering and management.

KC: They realized that there had to be in-line adjustments built onto the board to compensate for the drift of the components.

LF: Basically, yes. The problem, though, was that this didn't work particularly well either. What he had done, in effect, was tune to the performance of the master tape recorder - the AG-440 master. We proved with a recording oscillograph that feeding the [55 Hz tone] bursts into the master tape recorder caused the circuitry in there to go into an oscillation, so that the baseline of the signal would wobble drastically, and the audio jumped around at a very low frequency.

So this guy had been totally empirical, and tuned his circuit to that bouncing behavior. He just scoped it and caught the level. This worked until we changed the master tape recorder for the next system, and suddenly it stopped working. I could tell that this was getting us into trouble and I began to advocate using automatic gain control - control of the signal's variations. I kept an automatic gain control in development as long as I could. I had to stand up at review meetings and be told "Kill that, freeze the design the way it is and go on." Eventually I did as I was told.

A couple of years later I went back to see what had happened, and discovered that an engineer named Al Alcorn had been given the circuit to fix up, apparently by putting in automatic gain control and increasing the size of it from one board to two - in short, the things that I had wanted and tried to do.

KC: This is the Al Alcorn -

LF: - who co-founded Atari. And in fact, a year or two later I was sitting in front of his desk applying for a job. Al was looking for manufacturing engineers at that time. He didn't remember much about the tone detector, but it was certainly an embarrassing and instructive moment. That event and a lot of others have kept me from ever really feeling good about the wisdom of management.

At the end of 1971, I had been doing psychotherapy for a while and things were getting better. I took educational leave to finish up my degree. Since I was on the co-op work/study program, I was able to convince myself that my combination of class work and experience had put me five years ahead of my classmates; I had eaten up four of those years, from the end of '67 to the end of '71, so it was time to get my degree - which I did, in June of 1972.

The Pyramid System project was moved into the Videofile building, where Alcorn was, and I spent a few weeks commuting down to Videofile. Then I left, and Ampex basically collapsed. Everybody got laid off.


Meanwhile, in 1971, a guy I had known in my activism days came running up to me and said, "There's some people over in San Francisco that have a computer, and they're going to use it for non-profit activity." I decided to investigate. This was a group of four computer science students who had left Berkeley during the Cambodia crisis, and essentially dropped out of the system as it was; they had come to rest at a "warehouse project," or warehouse community, called Project One in San Francisco - where a group of people basically formed an unincorporated non-profit association and leased a warehouse. Within this Project One, these four guys created Computer Group One, and were trying to deliver computing power to people and organizations who lacked access - to non-profits, to radical groups, to social-action groups of every kind.

Through an inspired act of hustling carried out mostly by Pam Hardt, they managed to get an obsolete time- sharing computer - an XDS 940 - donated together with money to set it up and run it. This machine was serial number 4 and was donated by Transamerica Leasing; it had been down at SRI and apparently run a robot known as Shakey. We understood that name very well, because the 940 had a DMA channel which occupied a whole cabinet, and we went deep into it and found the terminators for the bus were wired to the wrong voltage - to four volts instead of eight, which put them square in the transition region. That would make anything connected to it "shaky." I think this 940 had also been used by Doug Engelbart, so it had had an interesting life.

They had also begged a time-shared BASIC and were starting out with time-sharing, which was ambitious to say the least. I signed on basically as chief engineer. I was to be trained in system maintenance by a guy named Al Montoya who lived at Project One - it was living and working space both, kind of like four stories of lofts - but Al disappeared the day it was installed and didn't turn up for months. When I complained about his leaving all he said was "Here I am." Meanwhile I had absolutely no tutelage and no source of any; for one thing those computers weren't all that common, there were only 58 ever produced including conversions of 930's and 9300's. The language we used for our information retrieval system, called QSPL, was only available on 940's or on IBM 360's which wasn't necessarily a time-sharing implementation.

KC: Now, at that point the computer had been moved from SRI to San Francisco?

LF: It had been retrieved. Transamerica Leasing had leased it to SRI, who obviously kept it for the requisite three years. At the end of the lease Transamerica didn't have another taker for it and chose to, in effect, donate it or long-term loan it to this non-profit, Resource One. So it had to be delivered in a couple of trucks; it was a row of about a dozen 19-inch relay rack cabinets, each one about two feet wide - 19 inches is the internal dimension - and so 24 feet of cabinetry. This machine also required 23 tons of air conditioning. We had to run our own power lines from the main power busses downstairs to keep the thing happy, you didn't just plug it into the wall. It had a fairly big line printer, a console that sat on a table, and a Model 35 Teletype for the console terminal. The multiple serial interface handled 8 lines, I think, and I had to design and build a rack of modems where we could collect, if I recall, four modem lines. It was usable by ASCII terminals, mostly Teletypes which we'd begged from Tymshare after their lease contracts had expired. They'd been used pretty heavily and when we got them they were kind of sticky.

I think that, to today's computer user, the most amazing thing would be the disk file we bought for it. This was a double width cabinet, so it was about four feet wide, and it had two pull-out drawers each of which would hold a 2314-style disk drive with a stack of 14-inch platters that were removable; you took a plastic dome cover and slipped it down over the pack, and you could detach the pack and pull it out. We got one with double density, so it was 58 megabytes. To fill just one of the two pull-outs cost us twenty thousand dollars.

KC: You say like a [IBM] 2314, but not in fact a 2314?

LF: It was manufactured by Control Data, I think. We worked with people from Systems Concepts, which was a mile away in San Francisco, and Fred Wright - I think - designed and supervised the construction of an interface which emulated the Data General Nova back panel, and let us plug in a Nova disk controller that we had bought. I did the mechanical design of the whole thing, and that and the interface got it working. The system also had a swapping drum, hundreds of fixed heads arranged around a big conical drum, which would spin up; centrifugal actuators would fling out and the drum would hop up into engagement with the heads, because it was tapered towards the top. But this drum was flaky - the surface was deteriorating, and nobody knew how to fix it. We had a whole bunch of problems! The Ampex TM-4 tape drives had gas thyratrons driving little solenoids that actuated pinch rollers, and if you missed releasing one of them you could snap your tape, because the tape was going in one direction and the other pinch roller would come in pulling in the opposite direction. Mostly it would just squeak and stay there, but it could snap the tape.

KC: These were capstan drives, not vacuum-column drives?

LF: No, they had tension arms to buffer the speed variations. The tape wound back and forth along these little roller guides in the arms and interleaving stationary guides. So the arms followed these arcs and their positions were sensed, and when they got too far out, the tape would be speeded up. So no vacuum columns. I muddled through somehow, and certainly it was all a good education in antique electronics. But I never really got trained strategically, to answer questions like: What is the structure? What is going on inside this computer? Those answers would have been distinctly useful, but they had to wait till later.


In 1972 we had the thing installed, and we embarked on writing an information retrieval system. Richard Greenblatt came through town at just that time and gave us a combination lecture and pep talk with a focus of, "Let's write an information retrieval system in one day." And he sketched out a free-form keyworded information retrieval system whose data could be arranged according to new index words on the fly. The easy way to do an information retrieval system is to determine what words you're going to use as indexes, put up a list, and have the user enter an item by choosing from the list. The hard thing is to make that list expandable in the middle of everything, so Greenblatt lectured on how to do this. I had brought a friend, a systems programmer named Efrem Lipkin, who had an appropriate background, and he took on the direction of the project. The final code was called ROGIRS, the Resource One Generalized Information Retrieval System.

We were going to use this for the benefit of the Bay Area switchboards - volunteer information and referral agencies. Anybody with a card file box and a telephone could set up as a switchboard on whatever topic or in whatever area, publish a free notice in the underground press, and be in business. The trouble was that, for the great majority of switchboards, the filing system existed only in one person's head. When that person got burned out and left, which was only a matter of time, another person would have to come in and start the filing system from scratch. It was quite inefficient.


Now, Resource One had been attending meetings of switchboards in San Francisco, because they had taken over the corporate shell of something called the San Francisco Switchboard, the other half of which had become the Haight-Ashbury Switchboard. We were going to give the switchboards a common filing system and enable them to share resources, thus the name Resource One. We took the better part of a year to get that retrieval system written and debugged, at which point we went back to the switchboards, and discovered that all our contacts had burned out and left and nobody remembered us - although one person in the meeting apparently remembered hearing about us. So for practical purposes we walked in cold, all smiles, proposing that each switchboard pay $150 a month to rent a teletype and a modem, so that they could then laboriously key in their files, and only then could their users get anything out of the system. Nobody knew how to deal with this, none of the switchboards had $150 a month to spend, and we were not really treated seriously.

So we then had a reasonably powerful, empty information retrieval system and went about looking for things to do with it. We talked to many people, including some librarians at the Bay Area Reference Council - which is the library of libraries; if your library doesn't have a book, the Bay Area Reference Council is the organization through which it gets the book from another library. The librarian understood what we wanted to do, but said "You have something a lot like a set of shelves with no books on it. Why don't you get some books on it and see what it looks like?" Efrem was inspired by this and had the idea to set up a public terminal in Berkeley, where he lived.

At that time, about June 1973, I had burned out from the stress of living and working at Project One. Efrem and I realized that, if we could establish a combination office and apartment in Berkeley, I could live there and disengage myself from the pressures of communal life, and Efrem wouldn't have to commute to San Francisco. We proposed this to Resource One and convinced them to finance it.


By August of 1973 we were ready to put the public terminal up. We had a possible location in a record store that was being run by the [UC Berkeley] Student Union, called the Leopold Stokowski Memorial Service Pavilion - later it became Leopold's Records. We went to the Student Senate and said "We'd like to put a computer terminal for an electronic bulletin board there," and they said "That sounds great, why don't you do it?" I built a foam plastic enclosure for a Teletype 33, with a clear vinyl flap that attached with Velcro, to muffle the sound of the print hammers. It had two ports in the front covered with overlapping vinyl flaps, like a cat door, so you could stick your hands in and use the keyboard. Now, Berkeley didn't have local phone service to San Francisco but some Oakland exchanges did, so we used an Oakland prefix number installed at the store - apparently it was a residential number and I still don't know exactly how this worked - to make one call per day; that is, in the morning we would dial up the phone and plug the handset into the modem, and have a free connection to the 940 in San Francisco all day. We were using an Omnitech modem, or something like that, which was good for 300 baud under the most favorable circumstances. I remember that the modem cost $300, which would be the equivalent of $600 to $900 today.


Then I started exploring the possibility of building a modem ourselves, and I was encouraged by Fred Wright and the folks at Systems Concepts, who said, "Well, essentially all you'll have to do...." These were MIT guys, who gave me that word "essentially," which really meant "and someone else will do all the hard work," like the MPD engineers. But I didn't know that; I just went ahead, tried it out, and hit upon something.

The first part of building a modem is to convert digital data to tones, and that's easy. The hard part comes when you have to detect the two tones and discriminate between them. In essence we have a phase-locked loop oscillator circuit, which produces a voltage that varies with the frequency of what it's locked to. Against this we need a reference that says, when the voltage crosses this line, the bit changes from a one to a zero. The hard part was setting that reference and keeping it set, because it would be perturbed by component tolerance, by temperature, and by any number of other factors.

KC: You were trying to draw a line down the middle of a graph, and if that line was any less than straight, you risked misinterpreting bits.

LF: This is, again, "how far off the ideal you can stray and still be safe." You can draw as straight a line as you want, but the signal in the graph is going to meander somewhat, jiggling back and forth all the time. You're dealing not only with short-term oscillation but with long-term drift. What I realized was that the modems of that day - and mostly of this day too - were all specified down to zero baud. That would be preferable for something like a burglar alarm, where you have a rented phone line, a modem and a switch, and if that switch ever opens you want to know about it at the other end. But that's not what we're using it for. We're using it for data communication, and that implies a minimum rate of transmission below which, who cares? So that became the first breakthrough.

The second insight came from the work I had been doing to survive. At this point I was consulting for my former bosses at Ampex, who had formed a little consulting company after the layoffs, and I was learning a lot about serial data transmission. What I learned among other things was that for asynchronous data transmission, in the absence of an error condition, the signal goes back to a one level between every character. (If it doesn't do that it's a break condition, which means "The line may have broken, look out.") I set up a floating reference, such that the "one" condition would always re-set the reference, and whatever voltage that "one" wanted to be was fine for the circuit; it would tend to drift slowly down towards "zero," but then the data signal would come in and bat it back up to a "one." It maintained very good referencing, unless you went into a line break situation and after a few seconds your data would turn to zero - which was appropriate, because a break is a break. Rather than referencing the signal to the external standard of a potentiometer, I was tying it to the last known "one" condition that had come down the line, which was much more accurate and more flexible. My goal was to run it off a cassette tape recorder, because that was a popular and inexpensive storage standard of the day, but the potential variations in speed and pitch meant you couldn't do that with regular modems. You could do that with this modem. I developed this in 1973, and we used it for the Berkeley terminal to the 940, and after further development it appeared as the Pennywhistle modem - the commercial kit modem - in 1976.

KC: Okay, the setup of the XDS 940 at the center of this ad hoc network of modems and teletypes was what became Community Memory. Now - are we to the Hazeltine VDT yet?

LF: Just about. We opened the public terminal at Leopold's somewhere between the 8th and 14th of August, 1973, I think it was the 8th. I planted an article in the Berkeley Barb that week, describing it and giving a political rationale for that model of distributed information, which still holds true with me. But in January of 1974 we decided to move the terminal to the Whole Earth Access store on Shattuck Avenue, which was then in its first year of operation, basically as a catalogue store for hippies and for communes. And we leased a Hazeltine 1500 CRT terminal capable of operation at a princely 300 baud - actually, the terminal was capable of much higher baud rates, but the modem wasn't. We connected it to the Pennywhistle prototype, which I don't think was called that yet, it was just "the modem."

Almost immediately, something happened that raised pivotal issues of public access and control - I wasn't present when this happened, but I was told about it. We had a service contract [for the terminal] and the service technician was working on it, and dropped the circuit board for the keyboard. The chips were ceramic packs in two clamshell halves, sort of baked together. The top of one of the chips popped off and you could see the actual die in there with the leads bonded to wires, and somebody who was looking over the tech's shoulder asked, "Isn't that going to be a problem?" And he said, "Oh no, it works." That experience soured my belief in contract maintenance, and we began talking about various ways to make a system like this develop and grow and survive in a public access environment. Efrem was in favor of armoring the equipment, to keep everybody out.


I took the opposite tack. My father had recently sent me a book called Tools for Conviviality, by Ivan Illich and published by Harper & Row. Illich was a former Jesuit rising star who got off the official track and began writing books about de-schooling society, that was in 1970, and went on to establish some little center that he worked from in Cuernavaca, Mexico. He had a perspective that admitted technology and yet was very much outside the industrial model of society. He described radio as a "convivial," as opposed to an "industrial" technology, and proceeded to describe basically the way I had learned radio, but from the standpoint of its penetration into the jungles of Central America. Two years after the introduction of radio in Central America, some people knew how to fix it. These people had always been there. They hadn't always known how to fix a radio, but the technology itself was sufficiently inviting and accessible to them that it catalyzed their inherent tendencies to learn. In other words, if you tried to mess around with it, it didn't just burn out right away. The tube might overheat, but it would survive and give you some warning that you had done something wrong. The possible set of interactions, between the person who was trying to discover the secrets of the technology and the technology itself, was quite different from the standard industrial interactive model, which could be summed up as "If you do the wrong thing, this will break, and God help you." So radio could and did, in effect, survive in that environment because it "grew up" a cohort of people around it who knew how to maintain and sustain it. And this showed me the direction to go in. You could do the same thing with computers as far as I was concerned.

KC: Although possibly the computer technology of the period was less forgiving than tube radio technology.

LF: Certainly; remember, I had backed away from it as a kid because it was clear it had to work at every clock cycle. Such reliability was unheard of to me, in light of factors like noise in circuits.


Well, I convened a discussion group around this whole concept, and announced it in Community Memory. Bob Marsh, whom I had known in college, turned up, and so did Ray Bruman - I think there were about five people in all. So we had several discussions about a computer appropriate for a public access environment, how it would be built, what it would be like; and my proposition, following Illich, was that a computer could only survive if it grew a computer club around itself. What is this computer like? How will it work? And we decided that the central aspect of the computer would be memory - random access memory chips were just then becoming available. Also, as part of the Resource One effort, I had heard from Don Lancaster who had written the September 1973 article in Radio Electronics about the TV Typewriter.

KC: Which he then expanded into the TV Typewriter Cookbook.

LF: Yes. In correspondence with him, and in a phone call from Resource One, I was trying to find out how to use TV Typewriters as terminals, because several people had come by and mentioned his article and said "You know, maybe you could use this."

KC: Okay, now. Just for my sake - in what way did this Hazeltine terminal _not_ resemble a TV Typewriter, or was it simply too expensive?

LF: $1500 or $1600 was far too expensive. The promise of a TV Typewriter was that you could take a TV set and connect it to this little box, which you could build yourself with a couple of hundred dollars - not even a couple of hundred, a hundred dollars worth of parts - and you would have a usable terminal. Technically, the structures were the same. It was an excellent exercise for someone who wanted to learn digital electronics, not just digital but general electronics, and _Radio Electronics_ got ten thousand paid responses for the two-dollar plans in the article whereas they might have been expecting twenty or thirty [responses]. So ten thousand people right away wanted to build it, and that was positive proof to me that we were watching the emergence of a convivial technology.

Having said that, I emphasize that the TV Typewriter was a very difficult thing to construct. There must have been quite a discrepancy between the number of people who ordered the plans and the number who actually built something. It used sequential memory, shift register memory, and was extremely tricky to build and debug because it used lots of little analog delay pulses. Bob Marsh had built one and he tried to expand it to improve its performance, and he never really got it working - but that's how he learned electronics! and that must have happened to quite a few people. Also, the TV Typewriter was not entirely usable as a terminal, because it was a paged display system; when you got to the end of one page, and that last character went up on the screen, you had only one-thirtieth of a second before the screen blanked and you saw the next [character]. So you were actually likely to be outrun by it.

KC: It didn't have any buffering for back scroll?

LF: No back scroll, that's right. I asked Don, "Why did you do it this way if it's supposed to be a computer terminal?" and he said, "It isn't supposed to be a computer terminal, people just want to put up characters on their TV sets." And he was right. So I had to think about why they wanted to, why "convivial technology" depended on doing something so simple. I concluded that people felt an urge to gain control over technologies that they knew were really important and were affecting, or would affect, their lives: computer technology, somewhat metaphorically in this case, and television. The TV Typewriter combined the two, and it was perfect. Lancaster was a significant figure as a result, so far as I was concerned. He had created a synergy between convivial technology and computer technology _and_ he had shown that there was a tremendous demand for this.


So I was picking into the design and working out some details, and I called Lancaster back, and he said, "I've got a new version that I'm working on and it uses random access memory." Now that idea planted a seed with me and I said, "Aha, here's a route into the convivial computer." Computers need random access memory, and once you installed RAM in the computer, you could use the same memory to run this terminal. I believe that that was, in effect, the genesis of the architecture of the personal computer, and if any academics want to make their careers by arguing this point, they're welcome to see me and I'll help them. My idea was to begin with a memory card and add a card that was built to put data into the memory, either from a keyboard or from a UART, and a third card to get information out of the memory and put it to the screen. Connecting the three cards implied a bus, so I defined a 44-pin bus structure that used Vector connectors, which we could get quite cheaply, and which was good for DMA transfers. That three-card set constituted a terminal. I defined what the terminal would do when it got various control characters, like, what happens when it gets a carriage return? - well, it sets the character counter to zero, but the line counter doesn't change. When it gets a line feed, the line counter would increment, and all these would be pointing to the memory. I put together a specification that I called "The Tom Swift Terminal, or a Convivial Cybernetic Device," and I mimeographed it and sold it for 25 cents. That was in the fall of 1974. But the common memory, and the bus structure, made this device potentially a lot more than a terminal. You could plug in a replacement front-end board, or maybe another display board, or an input editor board with more smarts and possibly even a microprocessor. Along the way, if you needed more memory, you could plug in more memory. Just as the boards plugged into the bus, I wanted to make sure you could plug the busses together to expand them. I never really finished designing the whole specification, but the goal was that the builder, the user, would control the rate at which the device grew upwards into an intelligent terminal and then into a computer _per se._ You'd just keep plugging! And that would be the realization of the convivial ideal.


In January of 1975, Resource One decided to shut down the 940. We had gotten permission to put [terminals] into the Co-op Markets, which would have meant a significant expansion, and we didn't trust that 940 to support a larger system, given that in certain ways it was marginal already. We weren't sure that the drum storage unit was going to survive, although apparently it did for many years after that. We weren't willing to risk the farm. We also didn't want to continue development along the dead-end line of QSPL, since what was that going to run on? Finally, I had worked on the [Data General] Nova and we were aware of the [DEC] PDP-11, both of which made it clear that minicomputers were the way to go. The Community Memory Project decided to split off from Resource One and go its own way, attempt to secure a minicomputer and move the software over to that, and come back with a system that was genuinely replicable. I think that, overall, it was a wise decision for the long term; for the short term, actually, it was not, but we weren't operating from a tactical perspective.


As you know, January 1975 also saw the publication of the Altair issue of the _Popular Electronics._ I was consulting and publishing out of my studio apartment, which was somehow full of a desk, a filing cabinet, a little work bench and a rack of shelves which stood over my heater grate. That was LGC Engineering - LGC stood for "Loving Grace Cybernetics," which was inspired by Richard Brautigan's poem "All Watched Over By Machines of Loving Grace" and had been tested out a little bit, it was hanging up on the wall at Efrem's place. It was intended that the Community Memory project would call itself Loving Grace Cybernetics, and when it incorporated in 1977 I believe it did use that name for a short while. We figured that LGC Engineering would be the future hardware arm of this.


Meanwhile, through these meetings in the fall of '74, I got re-acquainted with Bob Marsh and he was looking for something to do. Apparently his in-laws were helping to support his family, so he wasn't in desperate straits, but he was unemployed and casting about for something to build. He convinced me to go in on a garage with him, at 2465 Fourth Street in Berkeley, which cost $150 a month as I recall for 1,100 square feet. I set up my bench downstairs and he took this little loft office that was already in there, and put an air conditioner in it.

He and a friend of his, Gary Ingram, had access to a supply of center-cut walnut boards that were exactly eight inches high and about a quarter-inch thick - this had come from the Wisconsin woods through a third party. They were trying to decide on products to build with this walnut, and considering an electronic clock that would have a walnut case, trying to sell that through Bill Godbout or by mail order. That never went anywhere. They were also looking toward some improvement on the TV Typewriter, whether it was a redesign or just a kit to improve the existing one - that's when Don Lancaster told me "People just want to put characters up on TV."

Right then the Altair article came out. Bob showed it to me and said "Look at this picture on the front. It's clear it's a phony. Look at that lump on the side, that's not real." And the pictures of the boards that illustrated the article certainly didn't match what we could see on the front. So he said, "This thing has nothing in it, it's an empty box." From what we could tell, there was nothing much more to an Altair than the microprocessor interfaced to a series of connectors. Bob wanted to start building peripherals for it, so he and Gary formed Processor Technology Corporation.

I remained a consultant and never became an employee or part of the corporation, but I took on contract work drawing schematics - I knew how to draw a schematic properly after doing some drafting in 1967 - and writing the assembly manuals. I wrote the caution that basically said "If you are not an experienced kit builder, find someone who is and get them to help you," in effect warning that unless you apply convivial techniques to this you're not going to make it. I sub-contracted Jude Milhon - who was Efrem's girlfriend, and actually I met Efrem through her - to draw some little animated drawings of, say, a ceramic capacitor lifting a top hat and making a dance step, which demonstrated how you had to form the leads so that the capacitor would fit between the chips diagonally and still plug in.

Now, I did not do a lot of the engineering. On the 3P+S [Processor Technology port card] circuit, I figured out how to jumper the divider counters to get the various baud rates, although it's not easy to explain how that worked. Other than that, I may have made very slight modifications to some of the designs. But I was doing enough then to pick up a certain amount of money.

KC: We must be about to the first meeting of the Homebrew Computer Club - if I remember it happening during the first week of March '75 in Gordon French's garage.

LF: That's correct - I think it was March fifth. I can't recall whether Bob was doing any of this [Proc Tech] work prior to that, I don't think so, but he and I kept talking about the Altair.


Meanwhile, ever since Resource One days, on Wednesday nights I had been going down to potluck dinners at the Community Computer Center in Menlo Park, that was run by Bob Albrecht.

KC: Was that the same thing as People's Computer Company?

LF: Yes, I think sometimes they also called it People's Computer Center, PCC. On Menalto Avenue, on the Menlo Park-Palo Alto border, they had set up a couple of minicomputers running time-shared BASIC service to be a game parlor, in effect, for kids. What Bob Albrecht cared most about was kids and computers. He had written a book called My Computer Understands Me When I Speak BASIC, and some games for kids, and he published a newsletter that was also called People's Computer Company. There's a picture of Bob on the inside back cover of one of the earliest Whole Earth Catalogues, sitting teaching a bunch of kids to use a four-function calculator - which in those days was a big thing. He's got this brush cut, looks like a porcupine. The connection to the Whole Earth Catalogue lent a certain prestige because we all thought of Stewart Brand as the guy who knew exactly where it was at. So Bob set up the Center, and a guy named Fred Moore sort of attached himself to it; he would sign people up when they came in the door and was building a list, and he wanted to organize some hardware classes, although he knew next to nothing about hardware. The people at the Center tolerated him.

Meanwhile, one night Gordon French went to visit the Kaylor Electric Vehicle Shop a few doors down, and happened in on this place; so he ended up on Fred's list.


When the Altair was announced, Fred convinced Gordon to make his garage available, and then put out a call to the list. Thirty people responded - including Bob Marsh and me, because we drove down in a pick-up truck that I'd borrowed from a friend - and we all stood around looking at this unit.

That may have been the moment at which the personal computer became a convivial technology. You see, I had this particular Altair before the meeting; it was serial number 8 or something, it had been sent to People's Computer Company as a review copy, and they gave it to me. I took it over to Efrem's place and asked him what to make of it. Frankly, he considered it useless, and in a way he was right, since there was nothing to it but switches and lights. He kept it as a sculpture in his living room, on the same table with his guinea-pig cage, with its lights flashing to keep the guinea pigs company. I retrieved it and returned it to PCC, and it turned up as the centerpiece of the first Homebrew Computer Club meeting - by which time, apparently, it had a sort of critical mass around it. Thirty people stood and looked at it and started to tell each other what they knew about it. Steve Dompier was there and reported on the trip he had made to Albuquerque to try to get his Altair kit from MITS.

KC: Right. As I recall, he had sent MITS a tremendous amount of money and said "Send me one of everything," and back in due course, or undue course, came a letter from MITS that said "We don't _have_ one of everything."

LF: Exactly. I don't know whether he had to go down there to find that out, but he went down, and I don't think he was able to bring his kit back with him - he got it a little later. But he reported on what he had seen there, on what was in process, and he said it was a very small operation - which came as a surprise to a lot of people.

So the people in the room, including Steve Wozniak and Roger Melen, began to understand that maybe as a group we knew as much as these [MITS] guys, and that possibly the Altair wasn't even an action item. It changed our focus and we said, "First of all let's be in touch with each other, and sign this list, and tell each other what we're doing." The written record of that was reprinted and distributed to everybody as the first newsletter. I talked about Community Memory and the Tom Swift Terminal. Wozniak talked about the Breakout game he had done, and discussed the video terminal he was working on. Marty Spergel, who used to put together kits of parts for the _Popular Electronics_ articles and sell them, actually gave away an 8008 chip at that meeting, which was a very nice gesture.


Meanwhile, those of us who opened up the Altair box and looked inside weren't necessarily impressed. We discovered, for example, that out of four possible places for motherboards, only one was filled with a group of four sockets, and of the four sockets only two were occupied - one by the processor card, and one by the memory card. The memory card in turn had sockets for eight 256x4 static RAM chips, so the maximum capacity of the card was 1K bytes; but MITS only supplied two chips, so the stock Altair had 256 bytes of RAM, which was not enough to support the kind of programming that most people wanted to do. The processor card was nothing more than the 8080 processor driving the lines on the bus through buffers. They had installed separate data-in and data-out buses, which was really inefficient, because you never did data transfers both ways at the same time. That told me that whoever designed this didn't know what he was doing - didn't really understand what a computer bus was.

As we worked with [the Altair] we found problems that only confirmed that impression. For example, the Phase 1 and Phase 2 clocks were sitting right next to each other on the bus. Transmission line effects therefore caused coupling between the two, and the last place you want coupling is on a clock line.

The designer also used a positive-polarity Phase 2 clock - the critical one - and he should have used a negative- polarity clock, because TTL is much more effective on its pull-down than on its pull-up, and you have much more noise margin going from 5 volts down to 2.4 volts than you do when you're going from zero volts up to .8 volt. Many existing techniques could have been used to fix these problems, but none of them were, so it was a flaky design.

Also, the Altair needed everything. It needed I/O, it needed memory that worked. MITS offered a 1K dynamic memory card, not when the first machines were shipped but I think shortly thereafter, and they worked okay. Naturally, then as now, everybody wanted more memory and there was tremendous pent-up demand for a 4K board, so that people could run things like paper-tape BASIC. MITS began producing 4K dynamic memory boards from a different design, and they loaded all kinds of disk capacitors on them, but that's not all it takes to make it work. MITS never really got the 4K boards to work, but they sold them anyway. Almost all of them went out to customers and straight back to Albuquerque.

Bob Marsh didn't take long to figure out that he could build a board from the same 1K x 1 chips which I was using on the Tom Swift Terminal, [Intel] 2102's. Eight of the 1K-bit chips made 1K bytes, and then 4 ranks of those made 4K; and Bob did a good enough delay generator circuit, that board was produced as the 4KRA, I wrote the manuals, and Proc Tech was off and running.

At the same time the Homebrew Club - which wasn't called that yet - was growing amazingly. It met every two weeks. The first meeting was thirty people in Gordon French's garage; the second meeting was in the auditorium of the AI lab, John McCarthy's lab, at Stanford; and the third meeting was at the Peninsula School. At the third meeting there might have been a hundred and fifty people, although don't hold me to that, but I'm convinced there were more than a hundred, because they were going out into the halls to talk. Gordon French was trying to hold a lecture on computer science up front, and people were screaming that half the audience was outside - which I was too, because I wanted to meet these people. I already knew that somehow we had to impose some order on this process.

Now [Steve] Dompier in the meantime had gotten his kit and built it, and he set it up at the third meeting with a low frequency weather radio sitting on top of it. He plugged in the Altair and hand-entered the program [with the switches] and of course somebody kicked his plug out, and he had to toggle it in over again. We all wanted to know what was going to happen and he wouldn't tell us, he said, "Wait, wait, listen, wait."

Finally he pushed the Run button, and we heard music, and we could identify "Fool on the Hill" quite quickly. It was the strangest feeling we'd ever had, like "My God, where's the music coming from? What is this?" The radio was picking up the RF noise generated by the computer, and the tones could be modulated by programming the computer to do things at different rates. It was both absolutely brilliant and completely serendipitous, and more than anything it spoke to the quality of Dompier's intelligence, because it takes a particular kind of mind to catch this stuff. I'm not at all sure I ever would have figured it out.

When "Fool on the Hill" was over, the Altair began to play "Daisy." Now "Daisy" was the first song that had been played by a computer, I think in 1957 at Bell Labs. They accomplished basically the same thing, but with a much bigger computer and a speaker legitimately wired up to one of the bits. That tune was historically recognizable as the beginning of all computer music, which was why, in the movie _2001,_ the computer HAL is being gradually lobotomized and regresses to singing "Daisy." Dompier, I'm sure, had picked it up from there - as had most of us - and he had done precisely the right thing by invoking this to say that we claimed this history as our own, that we were as much pioneers as Bell had been, but that the computers singing for us were computers we could build and own ourselves. There was a tremendous message in that and we all responded. He got a standing ovation....

The conditions for use of material from the Analytical Engine require the inclusion of the following membership application form. Since CHAC has been quiet for some time, I strongly suggest checking on the CHAC website before submitting this form.
[Graham Seaman;]
     ____ Yes! Please enroll me in the Computer History Association
     of California for the next year. I enclose
     Individual membership: $25 on-line / $35 paper (circle)
     Corporate/institutional membership: $75 on-line / $85 paper
     Low-income/student membership: $15 on-line / $25 paper (circle)
      and send four issues of the ANALYTICAL ENGINE to
     ____ Internet address: ______________________________
     ____ CompuServe ID#: ________________________________
     ____ Other gateway: _________________________________
     My mailing address is:
     Name: _______________________________________________
     Company: ___________________________________________
     Suite, Box, Mail stop: ______________________________
     Street: _____________________________________________
     City: _____________________ Zip/postcode: ____________
     Country: _______________ Phone: _____________________
     ____ I will submit an article to the ANALYTICAL ENGINE. (Please
     refer to the guidelines for submission, above.)
     ____ I will help produce the ANALYTICAL ENGINE or do other work
     for the Association. Please contact me.