• 1968: IBM 360/67 running MTS, the Michigan Terminal System, in Fortran IV.

• 1970: PDP-8. In this class (IE 373) we wrote programs in PDP-8 machine language (i.e., not in assembler but in octal machine code), running the programs on a PDP-8 emulator that ran under MTS on the 360/67. I still have my PDP-8 Handbook ("Introduction to Programming").

• 1970: IBM 360/67. In this class (CCS 473, taught by Bernard Galler) we programmed in IBM 360 assembler, SNOBOL (using the classic Griswold, Poage, and Polonsky book), and BASIC (using an interpreter written by some graduate students for MTS the year before). Our "textbooks" for the IBM assembler were the classic "IBM Principle of Operations" and an IBM green card. Even though MTS was an interactive system, and terminals were starting to appear around campus, all our programs were done on punched cards (lots of IBM 029 card punches at the Computer Center), with line printer output.

• 1971: IBM 360/67. In this class (CCS 573, taught by Larry Flanigan) we wrote an operating system for the 360, running it as a virtual machine under MTS.

• 1970 & 1971: Burroughs B3500. I had a summer job these two years for Southwire Corp. in Carrollton, Georgia, working with the B3500. The first summer I was an operator, handling punched cards, magnetic tapes, printer output, and IBM card sorters and collators. The second summer I programmed the B3500 in both COBOL and Fortran. The Burroughs machine and its operating system were very nice to use and program.

• 1972-1973: PDP-8/L minicomputer at Michigan's McMath Hulbert Observatory. We had a 4 Kbyte PDP-8 for instrument control and data acquisition. We also wrote our own operating system for the PDP-8. MTS supported a nice PDP-8 assembler (since they used PDP-8s for data concentrators into the 360/67) so we wrote our programs in Ann Arbor, generated paper tape output, drove to the observatory (about 70 minutes away), and loaded in the program on a paper tape reader attached to a teletype. (High speed paper tape readers were expensive.) To avoid going back to Ann Arbor for simple fixes, you became very good at binary patching (putting in a jump to unused memory, and putting the new code there). This patching was done with the 12 front panel switches in octal.

• 1973-1975: Data General Nova minicomputer in assembler. The Nova was a 16-bit minicomputer, a competitor to the PDP-11. I was working for Singer's M & M Computer Division in Orange, California, and we developed RJE terminals (Remote Job Entry) using the DG minis. Our programs (basically stand-alone operating systems) had to fit in 4096 16-bit words of memory, and controlled all the devices (card reader, printer, card punch, and synchronous communications line). Each program was about 6,000-8,000 lines of assembler (3-4 boxes of punched cards).

• 1975-1982: At Kitt Peak Observatory most of my programming was in Forth on Varian 620/f minicomputers (another 16-bit minicomputer that was a competitor to the PDP-11). I also programmed a CDC 6400 in Fortran and Pascal and a Datacraft 24-bit minicomputer (later became a Harris/3) in Fortran. Around 1980 we got some PDP-11s and some VAXes and I developed the Forth system for both, and also brought up Unix on both. I also worked with RSX-11M on a PDP-11/70 at the University of Arizona Digital Image Analysis Laboratory, and did some work with VMS on the VAXes at Kitt Peak. Seeing how hard it was to use both RSX-11M and VMS, I became a Unix fan very quickly.

• 1982-1990: At Health Systems International I started on a 4-user PDP-11/23 running Venix (a Version 7 flavor of Unix from Venturcom). We then got an IBM PC XT (8086, 10 Mb hard disk) in February 1993 running MS-DOS with the Lattice C compiler. We later delivered a product on the PC XT running Venix and on the PC AT running Xenix.

  We were trying to work with IBM in delivering a product on the 16-bit Series/1 (yet another competitor to the PDP-11) running a little-known version of Unix that IBM provided for the Series/1: CP/IX ("Carrier Products Interactive Executive"). It was developed at Case Western Reserve University for IBM's Telecommunication Carrier Products Division (based in Princeton, NJ, to sell to the phone companies) and was based on Unix System III. I recall going to Princeton to evaluate their implementation, taking a Berkeley source tape with me. I spent a day getting vi and csh to run. Even though the Series/1 had separate instruction and data space (like the PDP-11; 64 Kbytes for each), the Series/1 instruction set was slightly less efficient than the PDP-11s, and vi used all 64 Kbytes of instruction space on the PDP-11. It took some creative hacking to get vi running on the Series/1. We did end up getting a Series/1 for development, but gave up on it shortly thereafter, when we discovered that a PC XT (later a PC AT) running Unix was just as fast, and much cheaper. Berkeley Unix on a VAX was also a much nicer program development environment than the Series/1.

  We replaced the PDP-11/23 with a VAX-11/750 running 4.2BSD, later upgrading this to a VAX-11/785. As I recall we were able to run around 16 users on a 4 Mbyte 750. Later we got a VAX-11/785 and then a VAX 8600 (upgrading the latter to an 8650). All our VAXes ran Berkeley Unix (I had seen how programmer unfriendly VMS was during my Kitt Peak days. Been there, done that.)

• AT&T 3b1 (aka, 7300 or Unix PC). This was a wonderful Unix system that unfortunately the marketing types at AT&T had no clue how to sell. It originally sold for $10,000; a few of us at HSI bought them when AT&T had a "fire sale" in 1987 for $2,495 (Motorola 68010 at 10 MHz, 2 Mbyte memory, 67 Mbyte disk, Unix System V Release 3, C compiler, Smart Software package).

• 1990-present: Sun workstations: SLC, ELC, SparcStation 4, and Ultra 5. When I wanted a workstation in 1990 Sun was the only one for less than $5,000. Most vendors have ignored the low end of the market.

Back to W. Richard Stevens' Home Page