Original from: ed-thelen.org/comp-hist/TheCompMusRep/TCMR-V08.html#Museum ------------------------------------------------------------------------ Highlights from ------------------------------------------------------------------------ The Computer Museum Report Volume 8 ---- Spring 1984 Contents of Highlights * The Director's Letter * The TX-0: Its Past and Present * The TX-0 at Lincoln Labs * The Move to MIT * The TX-0 at MIT * Debugging * Managing the TX-0 * Maintaining the TX-0 * Electronic Systems Lab Group * Cognitive Information Processing Group (CIPG) * Speech Research Laboratory of the Research Laboratory for Electronics * The Hackers * The Move to the Museum * Otis King's Pocket Calculator ------------------------------------------------------------------------ The Computer Museum is temporarily closed in preparation for its move to Boston. It will reopen at Museum Wharf in downtown Boston in fall 1984. For more information, call (617) 467-4036. ------------------------------------------------------------------------ The Director's Letter ------------------------------------------------------------------------ I'm often asked, "Will there be a lot of interactive exhibits in The Computer Museum?" I don't have a short answer. The long answer compares two exhibits: the TX-0, the first transistorized computer, one of the first computers used for interactive graphics, and a planned exhibition called "The Computer and the Image." Seeing the TX-0 exhibit-with its banks of switches, bulky Flexowriter, rows of toggle switches, wall of supplies, with its heavy steel chairs with peeling vinyl covers, tile floor and venetian blinds-sends the viewer back to the late fifties. And on days that John McKenzie, with his dapper bow tie, set of tools and complete machine log, is busily maintaining the machine, the visitor has the extra advantage of a demonstration or discussion of the machine's state. One of its demonstration programs of a mouse learning its way to cheese (or a martini) has been a favorite for 25 years-even though its graphics don't measure up to those on a handheld child's toy. For many, this seems to be involving enough, although not really "interactive." To take the next step, to make the TX-0 operable by the visitor, is to put the clock back: the machine fills a room, yet is less powerful than many programmable calculators, takes a half-hour to start and demands programming in its assembly language MACRO. This is an investment in time that most museum-goers don't make. But what about the rare person who feels that they must program the TX-0? For them, the machine is to be simulated on the Museum's VAX. Dan Klein from the Mellon Institute has put a number of the instruction sets for the classic machines, including the TX-0, on the VAX. The serious visitor can then have easy access to classic machines at a terminal in the new library of the Museum. Another way to experience the TX-0 in action is truly vicarious: watching a 1961 made-for-television film about the machine, how it was used, with demonstrations of many of the programs. Such classic films will be used with many of the exhibits on super computers and vacuum tube machines that took in the order of 100,000 watts to turn on. Even if we got one of these machines together, the Museum could never afford to run it. Nor does it make any sense when the same computing power is available on a machine that requires a hundredth of the electrical power. To experience the size and power decline of the machines through the generations brings home the point better than any textbook statement. Historic machines will only be one dimension of The Computer Museum when it moves to Boston. Several exhibition areas will emphasize contemporary computing with interactive exhibits that demystify the "black box." The first of these is "The Computer and the Image Gallery" now being planned by Oliver Strimpel, on leave from The Science Museum, London. The aim of the exhibition is to convey the full breadth of computer imaging, from computer-aided design to the simulation of galaxy evolution, from Landsat image-processing to computer-drawn animation. In explaining the concepts of computer graphics, interactive displays will be used. What better way to understand how resolution affects a picture than to alter the resolution yourself? Or stretch the contrast, or distort the image with a simple mathematical function? Computer graphics can also portray objects that do not exist in real life. Interaction allows the visitor to walk around them or zoom in on areas of interest. Simulations using computer graphics often rely on the choice of parameters by the user, who gets more involved by entering his own choice. The focus of The Computer Museum is not to create "interactive" exhibits, but to preserve and explain the scientific and technological history of computing in the most appropriate and exciting way that it can be done. Gwen Bell ------------------------------------------------------------------------ The TX-0: Its Past and Present ------------------------------------------------------------------------ Up and Running. For the third time in its history the TX-0 was fully operational, this time at The Computer Museum. It was built at Lincoln Laboratory in 1955 as an experimental computer to test transistors, and had its first birth in 1956. Then in 1958 it was dismantled and reborn at Massachusetts Institute of Technology (MIT) where it operated until the mid-sixties. The TX-0 fills a room, yet has less memory than many personal computers. And unlike today's personal computers, the TX-0 demands a skilled user to maintain full-scale operation. The machine is a good example of what computing was really like 20 years ago. Jack Dennis and John McKenzie were responsible for the TX-0 at MIT before its shutdown. This time around McKenzie was the operations manager and the first to witness it come back to life. He worked for months preparing the classic, 1955 computer for its debut at The Computer Museum. His efforts paid off when TX-0 alumni, Museum members and other computer buffs united on Sunday November 13th, to display their enthusiasm for keeping the artifact in working, order. For one day, those who had be vital to the development and day-to day operations of the TX-0 reminisced about the days when it was at the apex of computer technology. A series of events were held touching on all phases of the TX-0's past. The day's events were videotaped for the Museum's archives. Included in the program was a lecture by TX-0 alumni and MIT Professor Jack Dennis on the history of the machine, a luncheon for the alumni, and a hands-on demonstration of the star attraction-the TX-0. One alumni favorite was the mouse and maze program. Everyone focused on the mouse as he scurried across the cathode ray tube (CRT) screen to catch a piece of cheese. Old stories about the TX-0 abounded during the reunion with some alumni suggesting they should have brought cards for old times' sake, because bridge was often played during breaks in the TX-0 room. ------------------------------------------------------------------------ The TX-0 at Lincoln Labs ------------------------------------------------------------------------ The TX-0 and its original software were created at Lincoln Laboratory. Its original team included Wesley Clark as the logic designer and Ken Olsen as the engineer in charge of building the machine. Phil Peterson, Jack Gilmore, John Frankovich and Jim Forgie worked on logic design under Wes Clark. Bob Hudson and Chuck Norman were involved in the construction working for Ken Olsen in the engineering group. The following quotes are excerpts from discussions and presentations at the Museum's TX-0 alumni reunion held November 13th. Jack Gilmore: The racks of logic that became the TX-0 were used to test both transistor circuitry and one-half of the 65K memory for the TX-2 computer that was on the drawing boards. Ironically, when everything was put together it turned out to be a fascinating general purpose computer. Phil Peterson: Imagine when the power was first turned on. We had no other machine to talk to the TX-0. In order to communicate with the machine, the toggle switch panel allowed a sequence of bootstraps so that you could start a function, such as read-in the paper tape. The first thing we wrote was an octal assembler to read-in symbolic codes. Architecturally, the machine introduced the idea of micro-coding. You could hit the carry pulse having never done anything else. A random number generator that would make nice little patterns on the face of the CRT could be done with no instruction. Once the programmers got the idea of micro-programming they were only limited in their experiments by their imagination. Jack Gilmore: Wes Clark wrote an assembler called Hark. It was arranged so that the symbolic addresses could be any length symbolic string, so we could incorporate English-language words, making programs easier to read. We put together one of the first on-line operating systems that could input software in a very easy fashion. One of the most significant pieces of work was a brain-wave pattern recognition program. Because the TX-0 had 65K of memory and an analog-digital I/O device, brain wave information was brought in as a moving window display. We were trying to teach the machine to recognize the "sleeping spindle." And because the TX-0 was the first machine with a coupled light pen and oscilloscope, the work was done four to five times faster than without these features. Jim and Karma Forgie did voice recognition on the TX-0 before they did it on the TX-2. Because the program took 10-20 minutes to execute, Jim h -' the problem of finding out where was. We rigged up a technique so he could draw a flow chart, put it on celluloid over the CRT, and in his software, slow down or speed up the program from the toggle switches. He could literally watch the voice program running through its steps. He could trap it and see it at a particular iteration using the Flexowriter or the toggle switches. Most software was for testing the 65K memory. The complete cadre of software included a print program, Hark, the assembler, a technique for online assembling, a search program to search through memory and find things, a debugging routine, and the dynamic flow chart program. A fair amount of this initial software was not usable for the new 4K environment that the TX-0 had when it moved to MIT. ------------------------------------------------------------------------ The Move to MIT ------------------------------------------------------------------------ In July 1958, the TX-0 was taken out of Lincoln Laboratories and installed in room 26-248 at MIT. John McKenzie recorded that it took 100 days of work to get the machine up and running. The following is from a memorandum, dated July 23, 1958, to TX-0 users from Earle W Pughe, Jr. : This memo is written as an aide to those who wish to write programs for the computer before the computer is in operation. The TX-0 has 4096 words of magnetic core storage. The cycle time is six microseconds, thus each order normally will take twelve microseconds. The inputs consist of a direct typewriter and a photo-electric tape reader. The outputs are a typewriter, a paper tape punch and a display scope. Other inputs and outputs are the toggle switch register and indicator lights on the control console. Provision has been made for users to connect their own equipment to the computer. It is expected that in the future the TX-0 computer will have more orders and more memory. Every effort will be made not to obsolete existing programs as new features are added. However to help meet the objective of not obsoleting programs as the computer is modified, the unused bits of an instruction must be zeros. This restriction means that such tricks as shifting a word to change instructions will obsolete a program when changes are made to the computer. Bits "0" and "1" are now used for the instruction, bits "6" thru "17" are now used for the address and bits "2", "3", "4" and "5" must be zero for all orders except "operate." The Ad Hoc Committee on Experimental Computation (Chairman: Prof. J. E Reintjes) is the faculty group in charge of the computer and they have final decision as to who may use the computer. It is expected that with the cooperation of the users there will be a minimum of paper work in assigning computer time. Since the computer is to be used for experiments instead of for numerical computations, the blocks of assigned computer time will be considerably longer than with other types of computers. ------------------------------------------------------------------------ The Speech Research Group at MIT. Osamu Fujimura, Hiroya Fujisaka, John Heinz, Gordon Bell (with his hand over his mouth) and Professor Ken Stevens watch Pete Brady at the TX-0 console in 1959 at MIT ------------------------------------------------------------------------ ------------------------------------------------------------------------ The TX-0 at MIT ------------------------------------------------------------------------ Professor Jack Dennis: Because the TX-0 was created as a memory test computer, it had some peculiar characteristics. The size of the address for the TX-2 memory was 16 bits, while the TX-0 had an 18-bit word. How do you build a machine with a 16-bit address and an 18-bit word size? Since an ordinary single-address instruction format was used, only two bits were left for the operation code. How do you build a sensible machine code with just two bits? 1. You must be able to store information into memory locations. 2. You must be able to get information out of the memory, so one can operate on it in the central processing unit. The TX-0 does not have an instruction code "load." In the TX-0, one got information into the accumulator by clearing the accumulator and then executing an "add" instruction. 3. The third instruction of the TX-0 was transfer negative: transfer control to the location specified by the address. 4. The operate instruction was next. Anything not done by the other three kinds of instructions was done by operate instructions. The remaining 16 bits instead of referring to a memory location were simply a micro-coded extension of the operation code. One combination would cause a point to be displayed on the cathrode ray tube (CRT) whose coordinates were the right half of the accumulator and the left half of the accumulator. In the same instruction you could transform the contents of the accumulator so that it would cause (on the next repetition of. instruction) a different point to be plotted on the CRT Wesley Clark was a major force behind both computers. When asked what happened to the TX-1, his response was, "We don't build odd computers." So the plans for the TX-1 were scrapped just like the DEC PDP-3. Ben Gurly was responsible for engineering the display system for the TX-0, a unique piece of hardware that influenced his later design of the PDP-1 at DEC. The TX-0 was one of the earliest computers that allowed the operator to use the cathrode ray tube for interactive computation. In contrast, the displays on the Whirlwind were mostly used for recording information. The TX-0 display was used to show immediately the results of changes made to a program. In the fall of 1958 I had just finished my doctoral thesis and had been appointed instructor at MIT I also had just moved into an office in Building 26 near the TX-0. Not wishing to pursue further my doctoral investigations in operations research, I was open to new and interesting adventures. With a new computer down the hall, the hackery in my blood soon got me involved in its programs. This computer, unlike MIT's number cruncher, the 7090, had the feature of being intimate with its users. You could actually go up to the console and ask the machine to execute instructions and programs specifically for you. The display program, which generated interesting patterns, triggered immediate reactions to fix it up and try it again. If one was careful in choosing the number in the "live" register of the machine, you could cause some wonderful patterns. You could do this with a program consisting of a single instruction-repeated endlessly. Such informal interaction with a computer was completely new to the world. ------------------------------------------------------------------------ Debugging ------------------------------------------------------------------------ In the fifties a substance called FLIT was used regularly around the house to get rid of flies. Thomas Stockham and I called the debugging program we wrote for the TX-0, FLIT, which meant Flexowriter Interrogation Tape. It was a successer to UT3 written at Lincoln Laboratory and provided a medium for symbolic debugging. You could take the symbol table generated by the assembly program and load it into the debugger. The debugger could then talk to you about your program in terms of your symbolic addresses and symbolic instruction codes. FLIT allowed one to insert breakpoints in a program and then run it. The debugger would take over control whenever a breakpoint was reached, whereupon the user could interrogate the state of a program and decide to go on or not. The project to write FLIT was suggested by Professor Thomas Stockham who, perchance, shared my office in Building 26. We wanted FLIT to be a very interactive program, but we could not work with the cathrode ray display, perhaps because the character tables would take up too much memory. But more likely because many users would want to debug programs that used the display. Tom suggested that as soon as the typist had typed something that 1' was in error-something that would not make sense for any continuationthe program should tell the user about it. Tom invented an idea he called "hands lapping." Immediately upon typing an error, the program would type back a red question mark. This meant a lot of repairs to the Flexowriter because one would try to continue while it was typing back at you. Full duplex communication and displays have now eliminated that problem. About this time I heard of something called a "macro assembly" program and that Doug McIlroy had programmed one at Bell Labs. From this inspiration I wrote the program MACRO for the TX-0. This program was to lead to macro assemblers for the PDP-1 and other computers. MACRO turned out to be a large program. To debug it, it was necessary to use a simpler debugging tool than FLIT because FLIT took up too much memory, so I wrote a program called MicroFLIT FLIT and MicroFLIT were forerunners of debuggers written for other computers, including DDT (Digital ital Debugging Tape) written by Alex Kotok for the PDP-1. ------------------------------------------------------------------------ Managing the TX-0 ------------------------------------------------------------------------ After the TX-0 had been at MIT for about a year and half, I took over responsibility for the machine and immediately set about extending the machine's instruction code. Since we were not likely to afford the 65,000 registers the TX-2 had, we enlarged the op code to four bits and added an index register. The operate command was redesigned to provide more capability, including logical "and" and "or"; and more input/output orders. With these changes the TX-0 lost its original power to generate fancy patterns through repeated executions of a single instruction. Installing the new instruction set was a big undertaking. The machine was constantly in use by research staff and students. We made the changes by pulling one panel at a time during scheduled maintenance periods, and almost always had the machine back in operation on time. The correctness of the alterations had already been checked through simulation using a register transfer language to describe the new instruction set. John McKenzie has managed every move the TX-0 has made, and managed operation and maintenance of the machine while it was at MIT He can tell you what it was like to replace the switches on the TX-0 console. I recall that when the machine first arrived at MIT, several switches had special designations: one was labelled "Suppress Wes" and another was labelled "Dump Phil." These functions, doubtlessly referring to Wes Clark and Phil Peterson, are no longer present in the machine. In my time one principle user of the machine was Gordon Bell who was working with Professor Ken Stevens and Arthur House on speech recognition. Pattern recognition was of great interest also. Some of the people who worked on the TX-0 became heads of the Information Processing Technology Office (IPTO) of the Advanced Research Projects Agency of the government, where MIT has obtained lots of money to carry out research. One of the directors of IPTO was Ivan Sutherland, who created the program "Sketchpad" on the TX-2. This benchmark graphics program allowed a user to create sketches on the display by using a light pen. Using the TX-0 light pen, Ivan and Claude Shannon wrote a program that would search a maze. It would act like it was inside a cave and would decide how to move by following the walls. Larry Roberts, who also became director of IPTO, used the TX-0 for creating a kind of artificial intelligence program. His program recognized hand-drawn letters by learning from its experience. In 1961, when John McCarthy was advocating timesharing as a way to use computers effectively and DEC donated a PDP-1 computer to the Electrical Engineering Department, my attention shifted to building a time-sharing system around the new machine. TX-0 transistor. The first TX-0 transistors were in tubes to make it easier to test and replace them. The TX-0 had only 12 transistor failures, and almost every transistor that lasted more than 500 hours is still operational today. ------------------------------------------------------------------------ Maintaining the TX-0 ------------------------------------------------------------------------ John McKenzie: When the TX-0 was built, transistors that operated at a five megahertz speed were not available. Lincoln Lab put Philco surfacebarrier transistors, costing $40 each, into bottles that contained 10 transistors. These were designed to be tested in a "transistor-checker." Ken Olsen, Ben Gurley and other designers didn't know whether transistors were here to stay. The engineers thought they might have to replace transistors like they replaced vacuum tubes, or at least annually check them. With little deterioration after 10,000 hours, it was clear that these transistors were good. It wasn't worthwhile testing them anymore. No one cared and the industry was moving ahead to new products. At MIT, only a dozen unaccountable failures may have been due to transistors. Most transistor failures occured within 500 hours after installation. Otherwise they made it, and are still working today. Every time another feature was added to the machine, another power supply was added as a self-contained unit. The machine is cycled on in sequence and cycled off in sequence. You get the memory pulses before you turn on the read-write memory current. John McKenzie. who spent months revitalizing the TX-0 for its Computer Museum debut, enjoys watching the machine perform on TX-0 alumni day. ------------------------------------------------------------------------ Electronic Systems Lab Group ------------------------------------------------------------------------ Doug Ross: John Ward had only observed the art of programming on the Whirlwind. When the TX-0 came, John decided he should program. John Ward: I signed up and there I was in the room alone with the computer. I was terrified. Doug Ross: Earlier at Egeland Air Force we built an elementary mouse solving a maze problem on the 1103. So John and I did a mouse and maze program. I did the logic and John the display. John Ward: ...very slowly. There was no assembler. You had to figure out all the addresses yourself. The style of the program was reminiscent of Shannon's mouse that used relays. Doug Ross: It had more flexibility because we were able to use the light pen to place the mouse and either hide the three chunks of cheese or the three martinis. For MIT's centennial in 1961, CBS did some specials on the Institute. The CBS director said, "Gee, Westerns are so cut and dried couldn't you write a program for one?" And I was talked into it. The memory was used to keep, track of everything down to the actors' hands. The logic choreographed the movement of each object, hands, guns, glasses, doors, etc. A line of English script was written for each direction, even if it went wrong. That's how we got the loop sequence which was an actual error run. If you watch closely, the sheriff puts his gun in the robber's holster, and other strange things. Doug Ross. Seated at the TX-0's "L" shaped console, Ross explains how he and John Ward designed the Mouse and Maze program: "1 did the logic and John did the display." Doug Ross explains the flowchart for the logical choices in "Saga," the 1961 TX-0-written Western. Dit Morse: I've been asked if the error sequence was rigged. Well, it turns out that the CBS people were in the TX-0 room when the machine got into that loop. They saw what the programmer was doing and they grabbed that sucker so fast-they knew it was theater. The program's 13,000 line code was macro generated. One of the first and only programs that I wrote with a real deadline. CBS would not postpone the shooting under any circumstances. It took six calendar weeks to deliver six skits. ------------------------------------------------------------------------ Cognitive Information Processing Group (CIPG) ------------------------------------------------------------------------ Don Troxel: As a graduate student I used the TX-0 because I had alot of numbers to reduce statistically, and . was the best desk calculator around. People in our group started to use it because of the display capability. At CIPG under the late Sam Mason we measured reading speed. John Allen: The first speech synthesis by rules scheme introduced in England by Holmes, Mattingly and Scherm was first implemented on the TX-0. It made heavy use of this wonderful bank of switches to control the various parameters of that synthesis. Don Troxel: When Francis Li called me over to hear it, I expected it to have a Chinese accent, but it had an English one since that was where the rules were made. John Allen: We did experiments with pitch using the switches for control. The TX-0 and PDP-1 were used to start to build a reading machine for the blind. The character recognition part ran on the PDP-1 and the speech synthesis on the TX-0. The tenuous connection was often lashed together firmly enough so that we could read characters on the PDP-1 and have speech output on the TX-0. When John McKenzie let you turn the machine on, you were then part of the in group. One Saturday, a professor, who will go nameless, called me on the phone and said, "I just turned the TX-0 on and it won't go." I said, "Just put your hands on the console and don't do anything until I arrive." Fortunately he hadn't done anything disastrous. He just hadn't started up the clock sequence. Gordon Bell: Actually with improper clocks when you started you could ruin the core memory. John Allen: The price of the TX-0 was $3 million - from the development costs on the books at MIT Gordon Bell: That was a bargain because it led directly to the TX-2 and Digital Equipment's first products. Actor Jack Gilford played the role of the robber in a "shoot out with the sheriff." The climax of "Saga" written in 1961 by the TX-0 with the help of programmers Dit Morse and Doug Ross. ------------------------------------------------------------------------ Speech Research Laboratory of the Research Laboratory for Electronics ------------------------------------------------------------------------ Gordon Bell: I was a member of the research staff of Professor Kenneth Stevens' speech research laboratory The laboratory continues to train researchers and do research in analysis and synthesis of speech. Some colleagues who worked on the TX-0 included Arthur House, now at the Institute for Defense Analysis, Osamu Fujimura of Bell Labs, Hiroya Fujisaki, University of Tokyo, John Heinz, John Hopkins, Morris Halle, MIT, and Pete Brady. Speech was taken into the computer using a tape loop with sampling pulses on one tape channel. The audio (speech) signal was passed through a bank of 24 filters and read in via TX-0's Epsco analog-to-digital converter. The goal was to recognize the speech by analyzing the frequencies of the resulting acoustic input. The analysis was carried out by a technique we invented called analysis-by-synthesis; the computer posted a model of the speech and compared it with that to be analyzed by adjusting the model's parameters. Gathering vignettes. Steve Levey (left) who is writing a book on hackers, gathers tidbits from recollections of Electronic Systems Lab Group alumni Doug Ross (center) and Harrison (Dit) Morse. Reminiscing. Shag Graetz's first hands-on programming experience was at the TX-0 console, although he was a seasoned programmer before coming to MIT. ------------------------------------------------------------------------ The Hackers ------------------------------------------------------------------------ Alan Kotok: In the fall of 1958, I was one of the earliest of the undergraduate crew to come in. Jack gave a couple of introductory talks to the Tech model railroad crowd, and brought us over to demonstrate the TX-0. When we saw it, we said, "Oh, neat-there's all this time available." We negotiated with Earle Pughe and John McKenzie for time. They said if the faculty advisor was amenable, then we could use the machine without any supervision. Jack Dennis: As an undergraduate I wrote a large linear program on the Whirlwind to solve the transportation problem. After midnight, I could get my hands directly on the Whirlwind, and get scope postmortems all on my own. This led me to believe that informal direct programming by students was the way to work with machines. Then we formalized it on the TX-0. Dave Gross: I was a freshman at MIT in 1957 and got a tour of the new TX-0 computer room. In 1958 we, the model railroaders, discovered the TX-0. I was told that under no circumstances could I turn it on, since I was not an authorized user. The most elaborate program I wrote for the machine was a three by three matrix of dots that made a search. One night Alan Kotok and I had the idea that it would be awfully nice if you didn't have to run your program tape through the reader twice. So we wrote a program that put it on mag tape the first time with enough space for binary to be added. Alan Kotok: Before that no one had used the tape except to write from the beginning and fill it full. Here we wrote-and then left space along the way. Dave Gross: We tuned it to leave just the right amount of space. Alan Kotok: We put two load points on all tapes, with the utility at the beginning and then a point that allowed use at the end. We did anything to avoid having to punch another binary program on this Flexowriter that punches ten characters per second. Dave Gross: Alan, do you remember the expensive tape recorder program? You had your FM receiver here in the computer room and we said we'd hook up the audio to the A to D converter and write a program to record on that tape. Alan Kotok: That was digital recording more than 20 years ahead of its time. Dave Gross: It would write the whole tape as one long record. Play back through the accummulator created a whistle, so we used the scope's D/A converter fed back into the speaker that was under the console. Jack Dennis: Could you recognize Beethoven? Alan Kotok and Dave Gross: It wasn't bad, considering . . . Alan Kotok: After the PDP-1 arrived and before any of the fancy high speed links had been installed between the machines, the hackers of the day and I were contemplating how we could make use of both computers. We hooked up a serial line between the two with a buffered program to the typewriter. You could type a line at one machine and it would come out on the other. After we got it working, I said, "What can we do with this?" Someone said, "Play chess." Since some of us had been working on chess on the 7090, we got together a panel of chess players in the TX-0 room with a chess board. Some of us sat in the PDP-1 room with a chess board and waited for an unsuspecting chess player to walk down the hall and into the room. Some fairly gullible graduate student was enticed to play this great new PDP-1 chess program. Our victim typed his plays in. The group in the other room replied. It worked well for a while, but then there was confusion about one of the moves with an argument over the terminal. Alas, our victim smelled a rat and started for the door to the connecting TX-0 room. Gordon Bell: In the spring of 1960, 1 went out to DEC and bought some modules so that we could add a mag tape unit on the machine. Alan Kotok: And that took us into big time computing. Jack Dennis: I remember that my dream at the time was getting support for interactive programming on the TX-0, even though the one itty-bitty tape unit was the only bit of auxiliary storage we had. I was dreaming up schemes to keep peoples' files and images on this tape unit, so that one user could take the machine over from another, but that project was scuttled when the PDP-1 arrived in 1961. Then we started to use it to build a timesharing system. Shag Graetz: By 1961 this machine was a legend among programmers. I had been eased out of the nest at Harvard where I used the 704, with about three times this amount of equipment, that no ordinary programmer could ever use. I came to write a diagnostic program for the 906/2 tape drive-every bit the kludge that it appears to be. My first question was, "Who is the operator and how do I submit my programs?" Jack Dennis said, "This is it. What you see is what you get." The entire room of machinery was under the control of whoever was signed up to use it at the time. During the next academic year, I went to work for Doug Ross; the PDP-1 arrived and I moved over to work on it, where in our spare time we developed SPACEWAR! Deja vu. John McKenzie who was the technician on the TX-0 at MIT, once again readies the machine, but this time at The Computer Museum. ------------------------------------------------------------------------ The Move to the Museum ------------------------------------------------------------------------ John McKenzie: The TX-0's life came to an end when each of the labs got their very own computers. When I saw a note in the paper that Bob Everett was presenting the Whirlwind to the Smithsonian, then I thought that's the place for the TX-0. However, they weren't interested. A little bit later, I saw a short paragraph in a DECUS newsletter that DEC was starting a museum. And we said that was the way to go. Stan Schultz came down and we started to plan the move. We were about to move the machine on April 19th (a holiday), when the contract officer at MIT said, "Hey you can't give away this $3 million to a private individual." Everything came to a halt for two years. First it had to be offered to all military groups, then to all groups with government contracts, then various universities, then secondary schools, and finally to general services who could advertise it. On the first go-around DEC was outbid $2500 by an outfit in St. Louis that wanted the I/O. Then it was re-advertised. This time DEC was outbid $50 by a surplus dealer in Ohio. Roy Gould got busy on the phone and noted that it would cost a lot more than the price of $350 to move the machine. DEC gave them an extra $100 and took title to the machine. Stan Schultz: John Connally and I spent many hours labelling all the wires. John McKenzie: The dismantling took about a week, and then unfortunately it went into a warehouse for about two years. Stan Schultz: Initially we set up the processor and console, and it was on exhibit from the summer of 1979 until 1981. John Mckenzie: We never burned any bridges so that we could make it run again. Fifteen different power supplies were lost in the warehouse. Then we had to buy some new ones. But the CPU is pure. Twice in bringing up the machine I was stymied. Once with the core memory, and I called on Dick Best to do some circuit analysis and he got me out of that hole. Later on, in setting up the paper tape reader, Alan Kotok did some analysis and we made it work. It needs to be in a computer room environment with a cooler, steady temperature. Stan Schultz: While on exhibit, some people must have taken souvenir bottles from the console. When we let it be known that the machine was being brought up again, bottles would mysteriously appear on the console.