Log in

No account? Create an account
Previous Entry Share Next Entry

A memorial for a deceased colleague

On Friday I went to a memorial for an ex-colleague of mine, André Bensousson, who died earlier this year.

I first met André when I went to work on Multics as a GE employee in 1967 when it was a joint MIT-GE-Bell Labs project. The Bell Labs guys went off to build Unix (emasculated Multics) several years later. At that time GE was a part owner of Bull, the French computer company. They had sent André over for a year in 1966 to learn about Multics. This year got extended and except for about a 1/2 year he remained in the US associated with the Multics project (while GE's computer business was sold to Honeywell, Bull was nationalized, Honeywell's computer business was sold to Bull) until it was killed in 1986 and stayed a bit longer working on my next project before he returned to France. He remained a Bull employee during this whole period - probably a world's record for a foreign temporary assignment.

André was a methodical designer and analyst who always worked from first principles and derived an elegant solution to whatever problem he was working on. See the André Bensousson story on the Multics web site as an illustration of how he worked. I have rarely met anyone in the computer software business (and I have met many many people) who even came anywhere near to his style and methodology.

It was a low key gathering organized by a women who had among other things been with him for about a year while he was ill. There were some pictures of him including one when he was about 6 years old and another in his Algerian army uniform! as well as some more contemporary pictures. Among the attendees were about a dozen of my old colleagues (rsc included) and some of us went out to a delicious dinner at the Harvest Restaurant afterward.

  • 1
That's pretty remarkable, diagramming everything by hand and then coding it by hand. When I took my C++ course and my database course, I often coded by hand with pen or pencil, but I would invariably type it in to the terminal and compile it while it still had a gazillion bugs. Your former colleague sounds like he was a very impressive fellow.

He was. Among other talents, he was extraordinarily good at explaining things, clearly and accurately and in detail.

Note also the reference to a "calling sequence", not to mention the era when that took place: the code in question was in assembly language.

Actually it was written in PL/I. There was very little assembly language in Multics. It wasn't written in C since among other reasons C hadn't been invented yet (Dennis Ritchie and Ken Thompson were Bell Labs people on the project).

Oops. It was, of course, only after I wrote that comment that I realized that it might be wrong. I did know that C and Unix danced hand in hand into the proverbial limelight, and I thought I recalled reading that Unix was groundbreaking in being written in a high-level language, but I guess that was yet another thing it inherited from Multics. (Either that, or I'm confusing Unix with Doom.)

MIT people decided in 1965 to use PL/I for Multics. They had considered using MAD (Michigan Algorithm Decoder) which was used for CTSS (an MIT system) built in ~1962 but decided that a potentially commercial success language would be better. Burroughs wrote the OS for the B5000 in ~1961 in an Algol varient. I think there was a British system written in B (an ancester of C) in that time frame also.

  • 1