Back to the History of Project Delta
Recently I received e-mail from Bob Mader asking if I was the Clark Baker who worked on Project DELTA. I indicated that I was and was placed on an e-mail list of other deltoids (as we are now called). Most of the names were unfamiliar to me because they had joined the project long after I had left. A few of the original people have checked in and, based on their history, and one my dad wrote 25 years ago, I wanted to write down my memories of the events. This is more complete than it would normally be because I have had all the other stuff jog my memory.
Every summer in the sixties while growing up, I would visit my grandparents and uncle in Massachusetts. Usually my grandmother would take me to U Do It Electronics and I would pick out my one big present for the vacation. It was usually some electronics kit which would allow me to build a transistor radio or similar project. I was a good technician, and could always make the stuff work. Try as I might, I could never understand the descriptions of holes and electrons and how my project really worked. I was never able to design my own circuits.
In the summer of 1968 we moved from Delaware to Lewiston, New York and that fall I entered the 9th grade. I don't remember a lot about that year but the school owned a Minivac 601. The Minivac had lots of lights, switches, relays, a motorized turntable device, and a plugboard with wires for building all sorts of projects. This was my first exposure to digital (albeit relay) logic. I built lots of stuff with the Minivac. In general, I could design simple stuff myself, but I never understood how the complicated stuff worked (like how to make it play tic-tac-toe). I was later very impressed when a computer scientist at MIT built a computer out of tinker toys which plays tic-tac-toe. You can see it at the Computer Museum in Boston.
One day my 10th grade math teacher wrote a simple computer program in BASIC on the blackboard. I thought this was neat and wrote my own program that night and showed it to him. He corrected the mistake I had made (something like "LET A+B = C" instead of the other way around) and sometime after that we went to the teletype (an ASR 33 from Teletype Corp.), punched a paper tape of the program, made a long distance phone call to Union Carbide, put the phone in the acoustic coupler, and logged onto the Union Carbide time sharing system. Next we read in the paper tape (whose speed over typing helped reduce the phone bill), and ran the program. Over time I was allowed to use the computer a lot and wrote and ran many programs. I did my science fair project on the computer and won first place. The hard part was renting a terminal (the teacher wouldn't let his out of the school) for the day and getting access to a phone in the school where the science fair was being held. The judges didn't really seem to know what I had done. One other memory from the Lewport days was that I taught several people how to program and how to use the computer.
We moved back to Delaware just in time for the start of 11th grade at Brandywine High School. I was hooked on the computer and wanted to make sure that Delaware's computers were as good as what I was used to. Sure enough, the math department had a computer terminal and access to a B5500 time sharing system at the computer center of the University of Delaware. My dad remembers that this system didn't service the schools very well and that I came home each night complaining about how bad it was. Later the system became even worse as the school changed to using a 4 user timeshared PDP-8/L with no disk storage. The lady running all this (Teresa Green) told my math teacher (Mrs. Cook) that things were going to change again and get better and she was looking for interested students from each high school. She wanted to have a meeting of these students. Apparently I called Teresa and volunteered to host the meeting at my house.
I don't remember much about the meeting. I assume that George Robbins, an old friend of mine from before my two years in New York, attended. He was at the other high school in my school district (Concord High School). My younger brother (Eddie) was there. He was in 8th grade then. I remember a roomful of people from different high schools. I know that Teresa wanted a contact student or two from each participating high school. I don't know if she also envisioned a core operations group at the time.
As I got involved, I learned that the PDP-8/L was housed in a building called DICE (the Data Information Center for Education). Several school districts pooled their resources and ran this computer operation for report cards, school scheduling, and other administrative computing. It was an IBM shop with a 1401 computer, punch card input and output, a card sorter, a card interpreter (it would read the punches and print on the card what was punched), and a line printer. In addition, there were several card punches (operated by young women) and a verifying punch (important data was entered twice to reduce errors). In the back were some offices for the programmers and the computer operators (young and older men). Teresa had obtained an office for her computer.
Teresa's stuff was funded by DSAA (the Delaware Schools Auxiliary Association) -- a private foundation which provided seed money and building loans for Delaware schools. She called her project Project DELTA for Delaware's Total Approach to computing. As I understand it, it was a 5 year program in which a participating school would obtain its own computer terminal (TTY), dataset (modem), and telephone line, and pay a part of the operating expense of the timeshared computer and staff. Initially the school would pay a small fraction and by the end of 5 years, the school would be paying 100% of the expenses and the project would be self-sufficient. I believe DSAA paid for the computer. Schools with enough foresight to join in the first year got a better deal.
The summer of 1971 was spent planning for the arrival of Project DELTA's new DEC (Digital Equipment Corporation) PDP-11/20 (Programmed Data Processor) running the RSTS (Resource Sharing Time Sharing) operating system. I don't remember a lot of what we did that summer but I believe that Teresa, myself, George, Alan Hockberg, and Debbie Persolio went into DICE most days to work. The computer was delayed and wouldn't arrive until the start of school. We sorted ourselves out by what we liked doing or were good at. Debbie was in charge of the contact students, Alan was interested in how the hardware worked, George was interested in how the operating system worked and worked on (wrote, debugged, and maintained) the smaller, more short term computer programs. I deferred to George when it came to systems programming (though I did my share) and instead tackled the larger programming projects. Before the fall school year started we organized (probably Debbie and Teresa) a meeting of all the contact students (and maybe their parents so they would know where all their sons and daughters time was going).
The PDP-11 arrived along with Release 3 of RSTS. It was very buggy and Delaware was one of the first and largest installations. The high school students were probably much harder on the system than DEC's other users. I don't know what happened during the day, but George and I went over to DICE after school each day to see how things were going. With the system crashing every 45 minutes, Teresa knew she had a real problem on her hands. She got in touch with DEC and they sent one of the RSTS developers down to figure out what was wrong. He came with a 6 inch thick source code listing of RSTS! Back then, it was written in PDP-11 assembly code and assembled and linked on a PDP-10. The PDP-11 was too new and too small to support a development environment for its own operating system. In addition, the early versions of RSTS had to be configured for each particular installation -- there was no auto-config. We were certainly very curious about looking at the source listing and learning what we could about how RSTS worked. We also learned a little about how the DEC guy patched the operating system.
While we didn't yet have the knowledge to write our own code, we did learn how to create good bug reports when the system crashed. As I remember, DEC installed an auto crash dump routine so that when RSTS crashed it would dump all the registers and some of the stack, and reboot. We told DEC what it said and they would look it up in their source listings and try to figure out what happened, fix the code, generate a new system (and a new source listing), and send it to us to try. Sometimes the new systems were better and sometimes they weren't. The DEC people showed us lots of neat internal locations to look at. Over time, we wrote several programs which would print out these locations periodically. One day when I was home sick, Teresa loaned me a terminal. I was logged in all day watching various system resources, including the small buffer pool. I noticed that the number of available small buffers would go up and down, but when it got to about 5, the system would crash. We told DEC and they generated a new system with twice the number of small buffers. The crashes were reduced to one a day.
There wasn't a lot of opportunity to understand how the hardware worked or to make changes in it and Alan drifted away from the day to day operations. Teresa had various ideas of what kind of services we could provide to the schools (in addition to simply providing a computer terminal and time sharing services). One thing she wanted was a large library of canned applications. Some of these we got by typing in listings from books. Others came from DECUS, the Digital Equipment Computer Users Society. Still others came from projects similar to ours. Little of it was for sale and most had to be converted or modified because BASIC wasn't as standardized as some languages today. These programs could be used in high school courses (e.g. math, science, chemistry). Their successful use depended greatly on the teacher. I don't believe our library was as large or successful as Teresa wanted, but we did our best if a teacher asked for a particular program to be made to work.
Another service Teresa wanted to provide was the scoring of standardized tests. She believed that we could provide faster turnaround and better, more customized statistics than the existing, send it out to a big company to be scored, method. To implement test scoring, DICE leased a large optical scanning machine connected to a large card punch. We also purchased a slow, not very reliable card reader for the PDP-11. I believe that this was a piece of equipment which DEC purchased from a third party and interfaced (both hardware and device driver) to the PDP-11 and RSTS. I believe that initially Rick Satterthwaite was supposed to write the test scoring programs. It was a much larger task than any of us understood and, after a while, Rick's other activities (tennis) took precedence. I wrote a set of test scoring programs, and did a lot of test scoring. I remember the teachers not really knowing much about what they wanted in the way of statistics. Later, after we moved out of DICE, we got a new optical scanner from OpScan corporation. George interfaced it to the PDP-11 and wrote the device driver. I modified the test scoring software to work with it.
Assume you have a test with 5 sides (3 sheets of paper). On each side, the student has encoded a unique number (on some forms this is pre-coded). We scan all the forms in and sort by the number. In a perfect world, there would be 5 sets of data for each number. Usually many of the students were unable to code the same number 5 times (which makes you wonder about their ability to answer the rest of the questions). We had to go through stacks of papers and try to pull all the "bad" forms and match them up by hand. It was a real pain. Also, the new scanner was low cost which meant very slow. We got very backed up in test scoring. After a while we got out of the test scoring business. It was a good learning experience for me. So far, all the work had been volunteer. I believe I was paid $500 for the test scoring software, and that test scoring was done on a paid basis after that.
As part of services Teresa wanted to provide, a complex program called IMPRESS (Interdisciplinary Machine Processing for Research and Education in the Social Sciences) figured prominently. This program was written in BASIC at Dartmouth and contained many "modules" and databases (like the US census). Teresa took George and myself up to Dartmouth to talk to Edwin D. Meyers (the creator) and a programmer, and discuss how we could port IMPRESS to our little PDP-11. Dartmouth used some large mainframe (probably a Honeywell) with a large machine word (at least 32 bits) and their version of BASIC had many extensions. I don't believe they thought we were going to succeed. We discussed which modules to convert first, and settled on the two most used and important. As I remember, it was impossible to figure out their code so we wrote our stuff from scratch. I believe that my brother Ed did all the work and I made various programming suggestions. He called his version DEPRESS, because it was such a depressing program on which to work, and because you could say it was Delaware's version of IMPRESS. He got the two modules working and we obtained various databases from Dartmouth. As part of a summer job, he and I completely rewrote it, adding a third module, and called the result ECPRESS (Ed's and Clark's IMPRESS). In addition to being able to analyze existing databases, there was a provision for creating your own database, including using the optical scanner to read in data. I performed and analyzed a few surveys of my own, including one on drinking, seat belt use, and grades. Everyone I asked said they were in the top 3/5's of their class!
Another dataset analysis program Teresa wanted converted was called MiniTab. In the course of implementing this, I learned a lot of statistics (and learned to not really like statistics). By this point, we had moved to the EE department at U. of D. and I was able to ask various professors how to compute the statistics.
Teresa knew a lot about education and had followed the works of Seymour Papert (the LOGO language and turtle geometry). She wanted us to get turtles and a version of LOGO working on RSTS. We purchased some turtles from a small company (General Turtle) run by Seymour's brother and I made them work on RSTS. A turtle is a electric motor driven device which would leave "turtle tracks" on paper as it responded to computer instructions. In addition, the same box had a music board and we played a lot of music with it. Eventually Debbie had Prof. Papert come and give a talk at an IEEE meeting and we demonstrated the turtles working with LOGO on RSTS.
One day we obtained a complete dictionary in computer readable form. Ed had the bright idea to enter these contests which he saw in some crossword puzzle magazines. We entered one which we probably solved by a simple computer program and were declared a winner. To break the tie for first place, we had to send in some money and solve a tie-breaker puzzle. We wrote a better program which ran a little longer and sent in the answer. Again we were a winner! Again, we had to send in more money and solve an even harder tie-breaker which required an even larger program and even more run time. Eventually the problems got so difficult that we couldn't program and efficient enough solution to solve the problem. After a while, we gave up on these puzzles.
Teresa wanted to help the guidance councilors help students choose which colleges to consider. There was some program out there which ran on large IBM computers and probably cost a lot of money. I remember a name like CVIS (perhaps Career Vocational Information Service). I was to get a magtape of the database and make it work on RSTS. We didn't have a magtape drive at this time, and the one we had on order was delayed. This program worked with large bit vectors with each of about 5000 colleges represented as a bit. You started out with all colleges selected (i.e. all bits set). If you wanted only colleges in the northeast or south, I would "or" together the bit vectors of colleges in the northeast and south, and "and" the result into your running bit vector. Once the number of colleges dropped below 25, I would print out the selections. You were able to get information on any college. Once it was working, the program was easy and fun to use (assuming you wanted to go to college). To get the magtape reading software to work, I had to make a business trip to DEC and use a PDP-11 running RSTS with a magtape drive in the mill.
I worked for Delta the summer after 12th grade, and then went to MIT (fall 1972). Delta started at DICE and moved to U. of D. in the EE department (Dupont Hall or Evans Hall on the 2nd floor). Sometime after that, it moved to the 3rd floor. I came back and worked the summer of 1973 and summer of 1974. Each time, it took some time to get back into the swing of things. There were various EE and CS students involved (Linda Ruff), along with some high school students. The operating system kept changing and evolving, and I was learning lots of other stuff at MIT. By the time the summer of 1975 came around, I was too hooked into MIT and got a summer job up there (writing an optimizer in D for the D language which was a strongly-typed follow-on to C which never made it). I kept in touch with Teresa until the late 80's but never knew what happened to Project DELTA.
I haven't saved much from those days, but I still have pewter beer stein engraved "CMB DELTA '71 - '72".
While I don't remember this part very well my dad claims that he and the other parents would drive George, David Corbishly, and myself up to PMC (Penn Morton College) to take a FORTRAN programming course every Saturday (Fall 1970 or Spring 1971). Since George and I already knew some FORTRAN, we used the course mainly as a way to get access to more computer time on their IBM 1620. David and George used to live in the same development as I did. David's mom and my mom were friends.
Dave Robinson, a professor in the EE department at U. of D. played a large, behind the scenes role in Delta. He used to run a once a week seminar for high school students, making the PDP-8 (a classic 8), other hardware, and EE accounts on the University's time sharing systems available. I believe Dan Grim helped out here also.
RSTS had an extended version of the BASIC language. It was compiled into "push-pop" code which could be efficiently interpreted. Release 2 of RSTS didn't have the compiler done yet. I was told that certain programs were hand compiled into push-pop code and run on Release 2. The compiler and run time system were written by a consulting firm called EG&H. The letters were the last names of the founders. Tom Evans (went to MIT, worked in AI and wrote the Evan's geometric analogy program), Tom Griffiths (went to MIT), and Tim Hart (wrote the first garbage collector for LISP and one of the authors of the Lisp 1.5 manual (a classic from the MIT Press)). George and I wrote a push-pop disassembler and always wanted to write a push-pop assembler, but never did. Our names got back to the people at DEC and when EG&H heard I was going to MIT, they offered my a part-time job. I was employee #6 at EG&H and had a teletype in my dorm room and consulted for New England Telephone, General Turtle, and DEC. This gave me access to DEC and working at the mill. Later Ed went to MIT and also worked for DEC and EG&H. He married the daughter of one of the EG&H employees (she is an EE and went to work for DEC). EG&H is still in business in the same building in Lexington. DEC has changed somewhat and no longer has the mill.
Teresa took us to several DECUS meetings. I remember staying in the Parker House in Boston (creator of Parker house rolls). I don't know if we gave any papers at DECUS. A guy named David H. Ahl published some books of Basic programs. I believe that at least one DELTA program is in them. He later went on to publish Creative Computing.
I was always fascinated with programs which created mazes. There were several ones written in BASIC. No matter how often I studied the source listings, I could never figure out how they worked.
I remember trying to learn lots of computer languages. After BASIC, I learned FORTRAN (always being confused about the difference between the FORMAT of a statement and the FORMAT statement), and SNOBOL, and then ALGOL. George and I did a reasonable amount of hacking in ALGOL on the B5500 and later on the B6700. The systems programming on these computers was done in a dialect of ALGOL and we wanted to learn all about that. The computer center folks weren't too happy with our interest.
Dan Grim was always present but seemed to let us do our thing. I remember thinking that he was real smart (both because he was and because there is a huge gap between what a trained graduate student knows and a high school student). He would help us when we got stuck, but seemed to have his own set of things to hack on. He seemed to spend a lot of time with DELISA-Delaware's Extensible Lisp. I didn't learn lisp until I went to MIT, and then didn't really learn LISP until I started using the MIT LISP Machine.
We gave a presentation at Phillips Andover Academy (a private high school in New England). Probably it was on ECPRESS and probably they had a PDP-11 running RSTS. ECPRESS was publicized through DECUS and became very popular at a number of smaller colleges that were running PDP-11's. Several tried to recruit Ed for their Computer Science program.
What I don't remember very well is how well we provided a time sharing service. I know we tried to keep the system up and monitor phone usage. However, I don't remember talking to contact students and getting an understanding of how well the various schools were doing with the computer (either as part of a computer course, or running canned programs for other courses).
I wrote a drill program to help me with my Spanish vocabulary. I wrote a physics program which plotted waves on springs. In general, I tried to computerize everything. One of my college friends accused me of computerizing games and ruining them. A direct result of the move to NY and back and the heavy involvement with computers and Delta was that I changed from being a mediocre student to one who was much more disciplined and got better grades. Also, Ed and I stopped fighting with each other and instead worked together on the computer.
I got my S.B. from MIT in 1976 (with a thesis on square dancing) and my S.M. and E.E. degrees in 1980 (with a thesis on tools to aid in VLSI chip design and verification).
I worked for another year at MIT as Sponsored Research Staff and was invited to be a founder (one of 15) of Symbolics, Inc. It commercialized single user Lisp workstations which were developed and used at the MIT AI Lab. Symbolics grew to 1000 employees and $120 million per year in sales, and then crashed and burned. Around 1990 I worked at Kendall Square Research, providing CAD tool and verification support for their supercomputer (6 full custom chips and 4 ASICS). We shipped the KSR-1 and a follow-on, the KSR-1S with the largest shipment being a 160 processor system. The supercomputer market never panned out like we told our investors and KSR crashed and burned (helped by some accounting irregularities and an SEC investigation). In 1994 I joined my present company (Fujutsu Nexion) providing CAE and design verification support for a large, complex communications switch.
June 1998, Revision 4
Copyright 1998, Clark Baker
Back to the History of Project Delta