According to my offer letter, I was supposed to start at IBM sometime in June, 1976 (I’ve forgotten the exact date) and work on a compiler for the minicomputer that would be released as the IBM Series/1. However, I delayed reporting to work in favor of becoming engaged to Diane, and by the time I got to Boca Raton, the spot on the compiler team had been filled.
- First day at IBM: July 19, 1976, Boca Raton, Florida, General Products Division
My first job at IBM was on the team writing telecommunications code for the Series/1. I wasn’t enthralled, especially when it came to testing the code, because we had to short out the peripheral cards to load in our code, and I didn’t realize that there was only low voltage in the Series/1’s card case! And the idea of dealing with information that left the machine and might or might not return successfully was strange, too.
So after a few months, I made a deal with my management — if I would provide a complete unit test plan for the code I’d written, they’d transfer me to the Operating System group (just down the hall). They were happy to oblige, so I wrote the first of many productivity aids, one which read the structured assembler code and annotated it according to our unit test scheme (“Test Plan 7”). It took me a day or two to produce the test plan for all of my code, and away I went.
The Series/1 was an interesting beast; it had 64kb address spaces (we only used one in the first release of the OS) and not enough registers. It also had short and long forms of many instructions (the short form had, if I remember, 256 bytes of addressability from “here” — anything else took the long form). We were perpetually running out of room in the address space, and it struck me that it would be a good idea to have a peephole optimizer which would convert long form instructions to short form when possible. So I wrote it, learning PL/S in the process. And people loved it — it was in our “Quality Plan” almost immediately. That was the first time I really understood how much leverage smart tooling could provide.
I created quite a few other tools while I was in Series/1 development; I’m not sure why other people didn’t do the same thing, because I found that doing so always saved me time, and then providing the tools to my colleagues really provided a big payoff, at no extra effort. And I even got to learn new languages in the process (most of which I have, mercifully, forgotten), which I greatly enjoyed.
I mentioned that we were writing “structured assembler” code — this was in the days that using structured code instead of GOTO statements was considered an “Improved Programming Technique”. There were many “Improved Programming Techniques”, not all of which were used well.
In particular, we documented our code using a technique called HIPO — the documentation for HIPOs said, very clearly, that they were not used to document the logic of a module, just its overall function; however, our “Quality Plan” called for us to use HIPOs as flowcharts to document the logic. I found this silly (not to mention very tedious), and so I wrote my first memo at IBM: “HIPO: Threat or Menace” (playing off a National Lampoon cover), quoting the HIPO documentation to make my point. In those days, we actually hand-wrote memos, and the secretaries typed and distributed them — my department’s secretary tried very hard to get me to change the subject line, but I refused. In hindsight, the subject line probably didn’t help my case, but the HIPOs vanished.
It’s been a long time since anyone has had to type a memo for me; I can’t remember the last time I got a real hardcopy memo, for that matter. Email has taken the place of memos, and, for better or for worse, nobody else sees them before they go out to the world.
I wanted to find a way to reuse the subject line of my first memo in my last IBM email, but the joke wasn’t all that funny in 1976 and hasn’t improved with age. So my last email had a more personal subject: “33.7 years – that’s not too many”.
- Last day at IBM: March 31, 2010, San Jose, California, Business Transformation/IT
And you know what? I never did get to work on a compiler at IBM!
I didn’t know you worked on the Series1. I wrote GPIB code to take data from a jet engine and then control engine testing. I got to do quite a bit of work on the Series1 and System7 (predecessor) as well as a lot of other IBM products. Series1’s ran State Farm for years and were replaced just recently if memory serves me correcty.
My start at IBM was somewhat earlier than yours. I had about 19 years in by that time. You missed something by not getting to work on a compiler. Those were fun. I supported several in the field and got the opportunity to work on a compiler/emulator for DoD that took in Univac source, created runnable object code but actually ran on a IBM Mod 40. Eventually it got extended to the point that it could simulate the Univac code to make sure it would at least function.
You joined IBM the day after my birthday.
I remember, once upon a time, watching you do something and I was thinking, “Gee, if he ever has to do the same thing more than once, he writes a program to do it instead.”