Source for AI memos

[larsbrinkhoff]

MIT publications that we can potentially rebuild from source. (foo) means the source is available in the foo repository.

Document	Source	Type
AI Memo 147A	HUR; DDTMEM 188	TJ6
AI Memo 191	WP; BIBLIO 191A21
AI Memo 238	DEE; PRITS 42	TJ6
AI Memo 239	MB; HAKMEM 141	Plain text
AI Memo 240	(its) WP; 11SIM 240A38	TJ6
AI Memo 260	(its) WP; LOCK 260A17	PUB
AI Memo 261	(its) WP; PEEK 261A17	TJ6
AI Memo 267	MINSKY; DESIGN 76
AI Memo 285	WP; LINFND 285A43	TJ6
AI Memo 292	WP; UT 292A15
AI Memo 295	WP; LIGHT 295A13
AI Memo 300	WP; ARMDSG 300A1
AI Memo 301	WP; ARMCTL 301A1
AI Memo 307	LLOGO; MEMO >	Pub
AI Memo 313	JEA; LOGMAN >	TJ6
AI Memo 358	(its) TJ6; TJ6MEM >	TJ6
AI Memo 444	(unlambda) LMDOC; PAPER >	TJ6
AI Memo 447	(ToTS) ECC; EDOC >	R
AI Memo 452	(its) SCHEME; DOCUME >	TJ6
AI Memo 453	(its) SCHEME; META >	TJ6
AI Memo 495	(its-vault) BKPH; BRDF >	TJ6
AI Memo 519	(tots) RMS; EPAP 49	TJ6
AI Memo 519A	(tots) RMS; EPAP 4	Scribe
AI Memo 528	(unlambda) LMDOC; CADR >	TJ6
AI Memo 554	(emacs) info/emacs.mss	Scribe
AI Memo 555	(emacs) info/emacs.mss	Scribe
AI Memo 569	(its-vault) LLOGO; GC DOC
AI Memo 585	ASBELL; R PRIMER	R
AI Memo 623	(its-vault) MOON; AMBER >
AI Memo 643	(ToTS) RMS; LEP 59	Scribe
AI Memo 644	(klh10) supdup.mss (tots) RMS; SUPAP 6	Scribe
AI Memo 694 (unpublished)	(ToTS) SPACY; MEMO 15	Scribe
AI WP 16	WP; LINES 16W36
AI WP 25	VIS; VF25 1
AI WP 26	VIS; VF26 1
AI WP 27	VIS; VF27 1
AI WP 28	VIS; VF28 1
AI WP 29	VIS; VF29 1
AI WP 30	VIS; VF30 162
AI WP 32	VIS; VF32 1
AI WP 33	VIS; VF33 1
AI WP 52	FININ; WP52 2
AI WP 58	WP; FUN-FR 58W1
AI WP 59	WP; GT40 CONTEN (its) .INFO.; URUG MANUAL	TJ6
AI WP 63	WP; HYPOTH 63W1
AI WP 72	WP; XGP 72W1
AI WP 77	DHT; MAPPER INF	TJ6
AI WP 78	(its) WP; FED 78W1	TJ6
AI WP 80 (draft)	(unlambda) LMDOC; CONS >	TJ6
AI WP 87	WP; C.E. 87W1
AI WP 208	(mit/ai) LMWIND; CHOICE >	Bolio
AI WP 209	(mit/ai) LMWIND; OPERAT >	Bolio
AI WP 214	(its-vault) LLOGO; LMT 22	Bolio?
AI TR 350?	WP; C.E.TR 350T1
CSG memo 144	CLU; CLUPAP R	R
SYS.18.01	(its-archives) SYS.18.01.mss	Scribe
Dazzle Dart paper (preprint)	(its-vault) HAL; DAZZL DART	TJ6
Numerical Shape paper (preprint)	(its-vault) BKPH; SEM >	TJ6
Lisp Machine Manual	(unlambda) LMMAN;	Bolio
MTB-418	DLW2; BSGROE/BSGFNE
Mathlab Memo #3 "Introduction to ITS for the MACSYMA User"	TMQ; MC >	PUB
Mathlab Memo #3 "Introduction to ITS for the MACSYMA User"	TMQ; MCSRC >	Scribe

Many of these will have missing diagrams that were added after printing.

Some attempts at this: https://stuff.offog.org/its/remastered/

[larsbrinkhoff]

AIM-453, page 8, line goes beoynd right margin. Page 10 likewise, plus a column of characters to the upper right. Page 11, page 10 overflows?

[atsampson]

I've added more of these to the list above.

[larsbrinkhoff]

I added SYS.18.01 "A Machine-Independent MDL" which is a Scribe document.

[larsbrinkhoff]

SAIL; FAIL PUB, but now straying outside MIT.

[larsbrinkhoff]

Adding:
AI memos 191, 240, 260, 261, 285, 292, 295, 300, 301.
Working papers 16, 25-30, 32, 58, 59, 63, 67, 72, 78, 87.
Technical report 350?

[larsbrinkhoff]

Some working papers are called "Vision Flash" and reside in VIS; VF .

[atsampson]

Added AIM 307, the source location of which is mentioned in LLOGO RECENT.

[atsampson]

Added AIM 554/555, the Emacs manual, the source for which is on the TOPS-20 Emacs tape. The Emacs info file is also generated from the same source (i.e. it's a proto-Texinfo setup).

[eswenson1]

Would it make sense if I committed the binary for PUB, PUB2, and SCRIBE while trying to get the right pieces to build these from source?

[larsbrinkhoff]

Sure, binaries are ok when there's a program we want to use and it's hard or impossible to build from source. For example, I think it's ok to have F40 in only binary form. We had a FAIL binary for a while until I found the source on Saildart.

[atsampson]

Added Multics Technical Bulletin 418 as another potential one - see the list of MTBs.

[eswenson1]

By the way if any of you guys want an account on a publicly-available Multics system, let me know.

[larsbrinkhoff]

1161 adds four memos to this repository.

[larsbrinkhoff]

AI WP 214 - LLOGO; LMTDOC 22.

[larsbrinkhoff]

AI memo 623 - MOON; AMBER >

[larsbrinkhoff]

AI memo 569 - LLOGO; GC DOC

[larsbrinkhoff]

Trying to find the source of HAKMEM - AI memo 239. The memo says it's HAKMEM 140. I see these in backup listings:

MB     HAKMEM 17         2      ?/?/71
MB     HAKMEM 141        1      3/16/72 00:23:10.5
RCC    HAKMEM HAKMEM     1      2/9/72 23:10:13.5
ES     HAKMEM 25         1      2/27/73 21:11:10
ES     HAKMEM 26         1      3/1/73 20:54:23
ES     HAKMEM 30         5      9/24/73 19:15:57.5
ROWE   HAKMEM 29         6      11/15/75 23:20:09.5
ROWE   HAKMEM 57         3      11/22/75 22:43:15
ROWE   HAKMEM 58         3      12/17/75 00:40:09
ROWE   HAKMEM XGP        2      11/22/75 22:40:12.5
ROWE   HAKMEM 29         4      11/15/75 23:20:09.5
ROWE   HAKMEM 59         4      12/18/75 21:44:04

I have contacted the people I believe to be MB, ES, and ROWE.

[larsbrinkhoff]

• [x] ED doesn't have it.
• [ ] ES: haven't found contact info.
• [x] ROWE doesn't have it.
• [x] MB doesn't have it; says it was not TJ6.
• [ ] RCC: haven't found contact info.

[larsbrinkhoff]

MB wrote me:

I don’t have any file of HAKMEM. I cannot remember for absolute certain, but I am 99 percent sure that it was not in TJ6 or any text processing format. I believe it was a straight ASCII file. I think I consciously did it that way because I wanted to force a lot of formatting myself, manually. For instance, there are several places that have “ASCII art” diagrams of things. And I preferred to control line breaks and page breaks, to help convey the continuity or grouping of ideas. There are some illustrations that are hand drawn (such as the high performance video deflection amplifier schematic), some pen and ink plotter drawn, and some photocopied (such as the page from Knuth). All this was easier to manipulate when I had total control of the spacing.

[larsbrinkhoff]

LMWIND; OPERAT > is AI WP 209.

LMDOC; PROGR > is not AI memo 444.

[ams]

I think you mean LMWIND; OPERAT >.

[ams]

AI memo 444 is LMDOC; PAPER >

[larsbrinkhoff]

Thanks, yes it's LMWIND. 444 is already listed in the first comment. But which memo is PROGR?

[ams]

I don't know, I tried poking about and I recall reading this in a AIM.

[larsbrinkhoff]

AIM 238 is DEE; PRITS 42.

[larsbrinkhoff]

DEE; JDOCU 2 says the title is "ITS 2.0 Reference Manual" from May 1971. It seems to be written in the JUST text justifier language, DEE; JUST 114.

DEE; ITS WR ITE UP contains both "ITS 2.9 Reference Manual" and "ITS 3.2 Reference Manual".

[larsbrinkhoff]

Adding AIM 147A, which is close to HUR; DDTMEM 188 dated 1971-09-10.

@ericosman, is this your document?

[larsbrinkhoff]

Found MB; HAKMEM 141

[ericosman]

Yes, sounds like mine. My login for ITS was HUR. I documented DDT while I was working at the MIT AI lab during the summers from 1970 through 1974. I worked for Donald Eastlake who worked for Pat Winston. People I worked with: Ron Lebel, Mike Beeler, Richard Greenblatt, Bill Gosper, Richard Stallman,

[larsbrinkhoff]

Added the unpublished AI memo 694 "ITS internals" by @cstacy: SPACY; MEMO 15.

At the time of this writing, ITS has been in use for over a decade.
Until recently, most of the research at the MIT AI Lab was conducted
on the original ITS computer.  With the advent of such computers as
Lisp Machines, and the beginnings of radical devices like Connection
Machines, PDP-10 computers and their operating systems are becoming
obsolete tools.

This document is intended for the last people who will be maintaining
the ITS computers, and for historical purposes.  It might also be of
interest to the designers of future operating systems.

[larsbrinkhoff]

Added AI WP 52 "Tracking Wires on Printed Circuit Boards" by @finin, source code FININ; WP 2 timestamped 1973-12-06.

[larsbrinkhoff]

DHT; MAPPER INF is AI WP 77.

[larsbrinkhoff]

ECC; EDOC > is AI memo 447.

[larsbrinkhoff]

ASBELL; R PRIMER is AI memo 585.
CLU; CLUPAP R is CSG memo 144.

[larsbrinkhoff]

Added second file for 644: RMS; SUPAP 6. 643: LEP 59. 519 and 519A: apparently it was converted from TJ6 (RMS; EPAP 49) to Scribe (RMS; EPAP 4).

[larsbrinkhoff]

How to tell various formatter source code apart? Many use the RUNOFF syntax where a period first on a line is a command.

• If it says .xgp, it's TJ6.
• If it says .device xgp or some other device, it's PUB.
• R sources probably include R MACROS somewhere.

[larsbrinkhoff]

AI memo 519 "EMACS — The Extensible, Customizable, Self-Documenting Display Editor"

The first revision of the memo is written in TJ6. The second, 519A, was converted to Scribe.

Scan by @EricIO: https://archive.org/details/MITAIMemo519

New rendition by XGP emulator from RMS;EPAP 49. epap.pdf

[eswenson1]

The newly created XGP output has a 2022 date. I assume that is when you converted it to XGP format using TJ6 or Scribe? (Which formatted did you use?)

[qu1j0t3]

@larsbrinkhoff Looks good! I wonder if METAFONT parameters for device compensation (spread/thickening e.g.) were ever published for XGP; that might give a tiny hint as to device characteristics?

[larsbrinkhoff]

@eswenson1, it's the timestamp from the XGP file generated by TJ6. For some reason the ITS clock is way off! Or else it would have been February 2024.

I didn't convert to Scribe. The original, 519, was written in TJ6. It was converted (possibly also revised) in the 80s, resulting in AI memo 519A.

[larsbrinkhoff]

@qu1j0t3, intersting! I didn't even know that was a thing. What exactly is spread/thickening?

[qu1j0t3]

@larsbrinkhoff A notable feature of METAFONT is parametric adjustment of bitmaps to suit device characteristics. Every printing device has different amounts of "spread" or erosion through mechanical, chemical and optical processes. Without compensation, fonts can render too light or too dark. In fact a common complaint about Computer Modern Roman itself is that uncompensated outlines or bitmaps render "spindly" and faint on modern, precise devices like high resolution screens or offset lithography (imagesetter). The XGP undoubtedly has its unique characteristics here and since it is coëval with METAFONT usage, one might think that approximate METAFONT parameters for this were tested and published.

Quickly found reference to device parameters, for example: https://tug.org/TUGboat/tb08-2/tb18beet.pdf ; or see The METAFONTbook and CMR source files for full details.

[qu1j0t3]

All that said, your pdf looks pretty convincing to me :-)

Is it possible to get original XGP printouts from the personal archives of MIT people?

[larsbrinkhoff]

Found "Mathlab Memo #3" An Introduction to ITS for the MACSYMA User. TMQ; MC > is ~TJ6~PUB source, TMQ; MCSRC > is converted to Scribe.

[eswenson1]

When I try to convert that with TJ6, I get:

:tj6 ejs1;mc >
!XGPCO  unknown command  DSK:EJS1;MC 260(1,0)(0,1)
 DSK:EJS1;MC 260(1,0)(0,1)
.VAL 0; 7414>>JRST 16147

[larsbrinkhoff]

Right, it's looks a bit strange. I thought from the ^F font commands it was TJ6, but then maybe not. I don't know what it is!

[larsbrinkhoff]

I found this in another file, which looks much like the MC file. I believe this is PUB.

.device xgp
.!xgpcommand_";squish"
.font 1 "25vg kst"
.turn on "^F" for "%"
.margins(1000,1000,1000,1000)

[eswenson1]

Trying PUB:

ES>pub↑K!
*ejs1;mc >
INVALID INDEX NO. 1 FOR ARRAY MONTH
CALLED FROM 500134  LAST SAIL CALL AT 511104
?