STORIES ABOUT THE B5000 AND PEOPLE WHO WERE THERE By Richard Waychoff

December 9, 2012 3:56 PM

STORIES ABOUT THE B5000 AND PEOPLE WHO WERE THERE By Richard Waychoff [PDF of scanned document], or Text Version. [Via].

Entertaining memoir about the early days of computing, featuring several great anecdotes, notably one about college student Donald Knuth, hired by the Burroughs corporation to write an Algol Compiler on his summer vacation. Knuth, who needs no introduction to anyone interested in computing, describes the job in this video. (This appears to be the handwritten compiler source code and notes written by Knuth in motel rooms during the month-long drive in his Volkswagen to the Burroughs office in California.)

posted by JeffL (15 comments total)

Wow. Thank you!

Seriously hardcore geek porn there.
posted by pla at 4:33 PM on December 9, 2012

There is something fascinating, just comparing the format of those two documents. The Stories doc is right justified in Letter Gothic, which means it was probably produced on a Selectric Electronic Composer or a mainframe with ATS (Advanced Text System). The Knuth document is hand written on a form marked in the margin "HANO STOCK FORM 14111." It's some kind of coding form, and Knuth seems incapable of keeping his lines from drifting diagonally. After all the hours I spent using vertically aligned coding forms like this one, I feel like I was duped. My code was shoehorned into orthopedic wingtips while the pros like Knuth were dancing around barefoot.
posted by charlie don't surf at 5:26 PM on December 9, 2012 [2 favorites]

I miss MCP. I miss that Libra 185. I used MCP versions 48 to 51...

DMSII and disk arrays... tape libraries, and programmatic card decks.

I miss my days of being a Mainframer.
posted by PROD_TPSL at 5:41 PM on December 9, 2012 [1 favorite]

10 years ahead of my work on Array Processors think of them as the FPU for mini-computers of the 70's. 128 bit instruction word with a 38 bit floating point paths to the adders and multiplier one of the first parallel computing devices. Hand wired backplane hand tuned circuits made up of tens of thousands of discrete logic gates. Yes those days of bit kicking before lsi hardware and single chip CPUs existed was fun. Making machines run at speed reliably could involve rerouting clock wires to avoid inductance. Hand selecting RC pairs because tolerances were a bit sloppy those days. All in all it would take a couple of weeks to create the 12 Mflop machine depending on the amount of data memory and custom I/O boards (A to D converters and D to A display interfaces) the machine was priced from 100K up. Data memory was around 75K a Mw (mega word 38 bits). A single memory board cost more than my house did... The industry has come a long way 40 years. A functioning AP-120 was placed in a single chip in 1992 as a graduate project.
posted by pdxpogo at 5:56 PM on December 9, 2012 [4 favorites]

Just forwarded this to a buddy who did system programming on a B5000. Thanks!
posted by grimjeer at 6:26 PM on December 9, 2012

I just read five pages of this and it's the most electrifying thing I've read in a month. Holy crap.
posted by intermod at 8:41 PM on December 9, 2012

the handwritten compiler source code and notes written by Knuth in motel rooms during the month-long drive in his Volkswagen to the Burroughs office in California

I got an email from him, can you believe it? I tracked down one of the people he owed one of his famous cheques.

I kind of think that I know my way around a computer, right? But when I look at the work that some of these guys did, I feel like a carpenter who goes to work with a roll of duct tape and a rock. These guys will explain to each other how to write a compiler in _one goddamn day_, and then write their own assembler to bootstrap their hyperefficient compiler three weeks then, um, welp.

I don't know anything. I just tape things to other things, and then hit them with the rock until they're the right shape.
posted by mhoye at 8:57 PM on December 9, 2012 [7 favorites]

Holy crap is right, I read that from beginning to end, that was riveting. There has got to be a lot more of this story, so I checked around the web and there's nothing I can find that is accessible without an ACM membership. But I did find a tantalizing description of a 176 page self-published book by Waychoff, I Touched a B5000. Apparently this is a Burroughs slogan.
posted by charlie don't surf at 10:22 PM on December 9, 2012

Whereas I had thought that he was obnoxious as a college student, his increasing maturity clearly demonstrated to me that first impressions are usually pretty accurate.

Indeed.
posted by davejay at 10:43 PM on December 9, 2012 [1 favorite]

More technical info no the B5000 and successors (today's Xeon-based Unisys systems can STILL run programs written for the B5K in emulation!).

Here's an effort to write an open-source B5500 emulator.
posted by mrbill at 1:01 AM on December 10, 2012

I knew that it was possible to implement because Bob Barton told me so. He only forgot to tell me how... The meeting lasted two days. The focus was on how to solve the various individual problems that were posed by Algol. Most of the papers were presented by an obnoxious group of students from the University of Pennsylvania. When anyone else was presenting a paper they would laugh out loud, which clearly upset everyone else in the room...

Fascinating. Hilarious. Best computer memoir since 'Folklore'.
posted by ovvl at 6:57 AM on December 10, 2012

Useless trivia: I had always assumed that "Algol" was named after the Demon Star in the constellation Perseus...

It's actually named after the Persian mathematician Muḥammad ibn Mūsā al-Khwārizmī.
posted by ovvl at 7:02 AM on December 10, 2012

Note that the linked video is just one of a long series of autobiographical videos of Don Knuth including transcripts. If these historical topics interest you it is well worth watching or reading.
posted by mutagen at 11:41 AM on December 10, 2012

Ah memories! In the mid-80s I was a lab monkey and one of my duties was to backup a B5000 (among others) to 9-track magtape.

And every week I had to move 80+ tapes to offsite storage (a local bank's vault) wheeling strongboxes through campus on a handcart -- that was no fun at all.
posted by phliar at 6:06 PM on December 10, 2012

My plan for writing the compiler was to write a kernel compiler in Algol, then play computer with our flowcharts and generate the object code that our eventual compiler would have generated.

Holy crap. His original plan was to bootstrap the compiler by writing it and then compiling it in his head! They didn't end up doing that, but geez, to even consider it is seriously badass.
posted by jcreigh at 10:28 PM on December 11, 2012