Bob Saunders Interview: 9 February 2003
Technology historian Derek Peschel submits notes from his first of two visits to Bob Saunders. Bob was involved in the TX-0 project at MIT, in the late 1950s.
Author's marginal comments are set in italic text and marked by square brackets. Typos solely editor's responsibility.
9 February 2003
These top comments, plus a few in chronology (Bob Saunders next to MACRO II, dotted line) added 2003-03-30. Details about order of events could be wrong, but technical facts probably OK.
Visit with Bob Saunders --- Feb. 9 2003 --- Notes written afterward, same night.
(Initial date of Feb 9 and first correction to Mar 9 crossed out repeatedly with sidenote, "what was I thinking?" [crossed out] and "Japanese vs. US date order stupid brainfart + Mac not using leading 0s (2-9-03)" --ed.)
He asked me why I had come --- Programs don't get better automatically with time --- Bob Supnik/SIMH/PDP-1 --- Computer History Museum --- Borrowed folders --- They contained TX-0 listings by mistake --- Supnik gave me Bob's address.
[ Kotok said didn't know about it / I looked up his address. ]
He told me about the TX-0, testing large memory, small one added, large one taken away. Machine moved to Bldg. 26 (3rd floor?) which 704 was on ground floor of --- Only became usable after UT-3 was written then MACRO and MicroFlit --- Flit ran on larger-mem. along with some other progs. but Bob couldn't tell me any more, hadn't heard of.
[ Bob was glad not to have to use assemblers again --- skeptical of the HLL MPE was in --- "If you try to do it yourself, the compiler will kick your ass," said HP. ]
Expedite --- Jack Dennis as leader of TMRC "gang" --- A few other programmers like Pete Samson, I think Bob mentioned.
[ Basic interface (or "dialog" as Bob called it) stayed unchanged from UT3 on. He didn't know about UT1 or 2 or TX-1. ]
TX-0 had live register added first, then index --- I don't know when memory went from 4 to 8k --- New instruction set was tested using an interp.* (no virtualizing, just a simulator) which buffered tape since otherwise the reader wouldn't have run at full speed---This required extra deletes at the end of tape so reader wouldn't run out before buffer emptied.
* again when PDP-1 arrived, cross-assembler & emulator (?)
[ So Flit shrank to microflit, some features added back in MicroFlit S (I think) ]
[ FRAP required a big loader which was one reason MACRO was ported. Bob didn't know why it took longer for MIDAS to be ported ]
Bob didn't recall extend mode so perhaps during his time, the PDP-1 was still relatively small.
We went downstairs, Bob showed me docs, told me what he did after MIT---I told him about Data Discs & tour of Stanford CS dept. --- Check is Librascope the giant disk they have at the end of the hall now?
Then Bob showed his MPE terminal. real-estate software. SYSGEN and NMMOD --- we talked about HP's HLL. Burroughs 5000 Algol --- Bob asked why they didn't use an interpreter to create the first Algol compiler. --- Machine room, how to get a cheap HP system if you work for the company.
More about Bob's thesis, tape drive tests, & how he ended up in WA by falling in love. He said he never used TX-2 or Whirlwind w/someone from here. Later he said he also didn't know of the tic-tac-toe tape.
Back upstairs, I showed him Computer Museum docs. most of which were after he left MIT --- we talked about the informal process for tracking changes in the assembler/debugger/hardware, etc. --- "The word got around," people either propagated deps (i.e. assembler ---> debugger) or made sure their progs would work, but I pointed out the PDP-1 had undetectable config changes like HW mult. We also talked about why MIT encouraged style it did: continuous criticism, loosely (I'd say arbitrarily, but the results worked) defined prog. style, choice of features up to author, (I was thinking MIT encouraged thinking about features in some way, but no).
[ example in tx-0 code: load const to signify op in console lights, wait for input, sub. constant 2-inst. shift/test loop. converting to dispatch addr. from table, jump. ]
[ frequent copying of code incl. trick constructs. recognition of creation of same. ]
Bob Saunders p.2
I asked if there was a critical point when M/L progs. became hard to maintain, since Bob said TX-0 & PDP-1 progs were relatively bug-free, even MIDAS* desp. GC bug that Bernie Cosell told me he fixed.
* turns your programs to gold
Bob said foreground/background or other mult. processes / (perhaps also fancy I/O) but Arabic data processing threads makes that look easy --- disc. of combination of RTL for text mode, LTR for numbers, initial/redial/terminal <--- forms, ligatures, calendar, prayer times, dir. to mecca.
[ find alg. for calendar computations (how desk cals are printed) ]
[ I said I thought a priest had to see the new moon... "Well some one can see it and show it to a priest." Just a silly quote, I don't know if it's true. But since you can't know when the moon will be seen, printed cals are only a prediction. Why don't they adopt a new standard as they did with sunset? (astron. twilight, no messing around w/ threads) ]
Bob told story of kludges to deal with badly-formed ANSI tape labels, also story of filesystem corruption uncovering bug in emulated MPE firmware move inst.
[ HP did same as IBM w/ AS/400: change internal architecture invisibly ]
Also pointed out sec. on information carrying macros in PDP-1 MIDAS memo. (Appendix II) --- I think there may have been a TX-0 version of that memo, but Bob doesn't have it.
[ Bob thinks he wrote Memo PDP-3 hec. of style-it has no name ]
Hardware behind chess hack was Bob's doing. Also maybe SW that buffered a line of text before sending. Both machines used same logic levels so HW was relatively easy.
He mentioned Steve "Slug" Russell, but I forget why --- Chess hack? Weekend port of MACRO, maybe (this disc. was just after disc. of FRAP and DECAL)
Author of Expensive whatever?
[ DECAL exp parser is simpleminded---though it does offer op. syntax. DECAL is a later sys that included FRAP. ]
We talked about Hackers, errors therein and Steve Levy, (why did he write the book? I still don't know---Bob didn't either), the way stories spread & are changed, and the Hacker's Dict. Bob didn't know the story of PDP-1 HW change that broke DECAL but not MIDAS.
Spacewar --- Bob ported it to HP2100 w/point-plot display, HP sold for a while then Tek 545 scope when display stopped working --- Explained why Spacewar never got commercialized, which people have asked him: no cheap proc & mem, no controller to do vector gfx on raster CRT (vector CRT $30,000 and that's in 1960 dollars!)
[ I don't think he mentioned anything else about commercialization but there may have been a comment relating to that & bugs... or perhaps I'm remembering something I expected Bob to say halfway through a sentence. ]
A few other Bob Saunders notes---p.3
Bob was not entirely impressed --- as in "Oh, God" in "ugh" tone of voice --- when I told him about the machines that are being emulated: CDC 6000, LISP machines, etc.
If you're in Mecca, the dir to face is toward the mosque. If in the mosque, toward the stone ("The whole point of the thing.")
Me. "Generally non-Muslims can't get in, right?"
Bob: "There's nothing general about it---that's one of the rules of the game."
Me: "But haven't people tried to sneak in?"
(disc. of diff colored residence permits.)
[ Why is there no rail society in Saudi Arabia? ]
So as not to leave Ray Stricklin out of this---
Going we talked about the transmarticle system (of Fantastic Planet) C-Net for the Amiga, the advantages of BBSs (no broken clients as w/USENET threading). MCI terminal codes, TWENEX.ORG's BBOARD being proto-USENET ("watch when more people come")
Why did [Ivan Sutherland's] Sketchpad arise, so far ahead of everyone else?
Coming back we talked about MIT people & more about why env't was so productive* business that sprang up near MIT, stanford being a forced imitation, Microsoft being a parody of the culture, Linux as a new version of env't w/ same hyperfocused nerds --- the sort who want the alg. for prayer times, not the tables (as Ray pointed out)
* weak were weeded out. I think bus. were effect, not cause of innov: SAIL was formed by McCarthy.
We also discussed Expensive Typewriter, Planetarium, Desk Calc.
People asked about Bob & TMRC, wanted to find out about "ancient machine[ry?]" but Bob couldn't say what --- He has to check his e-mail.
Who's who: (written at end of the meeting. Copied here later)
Steve "Slug" Russell (co-author of Spacewar)
in 1992 was in Palo Alto, may be in san Jose now
Morton D. Cohen (used Whirlwind---Bob didn't know any other TMRC members who did)
worked for the Army (chem. weapons) after MIT
then I have "grad." --- grad student? graduated?
Steve Piner (author of Expensive Typewriter)
Bob may have his e-mail address