Classify tape images

[larsbrinkhoff]

Add a tool to classify tape images, especially those formats in use at MIT.

• ITS DUMP format.
• TOPS-20 DUMPER.
• TOPS-10: BACKUP and FAILSAFE.
• Skip optional ANSI label.
• Unix: tar, dump, cpio.
• Lispm tapes.
• VMS BACKUP.

CC @eswenson1 @ams

[larsbrinkhoff]

For PDP-10 tapes, we'll also have to consider whether they are 9 or 7 track, little or big endian, etc.

[ams]

Add a tool to classify tape images, especially those formats in use at MIT.

• ITS DUMP format.
• TOPS-20 DUMPER.
• Skip optional ANSI label.
• Unix tar.
• Lispm tapes.

FWIW, there are multiple formats for LispM tapes; Symbolics and MIT.

[larsbrinkhoff]

there are multiple formats for LispM tapes; Symbolics and MIT.

I can see that. I see something beginning with some ASCII text strings like

PRELUDE
VERSION 3
TAPE-SYSTEM-VERSION 429
LMFS-VERSION 428.1 RELEASED

Is that the (or a) Symbolics format? I think those are tapes from REAGAN.

And then there's something more binary looking.

[larsbrinkhoff]

I also see VMS BACKUP tapes. Hoping this will help: https://github.com/kkaempf/vmsbackup

[larsbrinkhoff]

I have some code up on a branch called lars/classify-tape. So far I try to detect:

• Big or little endian record length.
• SIMH or E11 tape format (odd record padded or not).
• For PDP-10 tapes, if they are 7 or 9 track.
• ITS DUMP.
• TOPS-20 DUMPER.
• Symbolics LMFS dump.
• VMS BACKUP.
• Unix tar - several versions; old/new, 16/32-bit, little/big endian.
• Unix dump.
• MS-DOS FAT file system(?!)

I see some unclassified tapes that look liike Unix or Lispm. The vast majority of tapes are ITS or TOPS-20, but there are quite a few Unix, Lispm, and VMS tapes that are as of yet totally unexplored.

[larsbrinkhoff]

Detection schemes range from heuristic to hacky. I mainly look at the first record, skipping any ANSI label if present.

• PDP-10 records should be multiples of 5 or 6 frames (9 and 7 track, respectively).
• ITS likes to write records of 1024 words, but may be smaller. A DUMP tape begins with a word -4,,0 or -4,,2.
• TOPS-20 DUMPER uses records of 518 words exclusively, so that's a good indication.
• TOPS-20 install media starts with an .EXE file; the first word has 1776 in the left half.
• Symbolics LMFS dump tapes have a bunch of ASCII text metadats; I look for the string "TAPE-SYSTEM-VERSION"
• Unix tar has the string "ustar" at offset 256 (decimal) in the first record. (There are other tar formats.)
• Unix dump has the magic number 60011 at offset 18, or 60012 at offset 24 (V7 and 32-bit Unix, respectively).

[larsbrinkhoff]

Hello @romkey,

I'm working on classifying tape image files from MIT's "Tapes o' Tech Square" collection of backups. Some of the files seem to be MS-DOS floppy images, or at least some kind of FAT file system. I was surprised to see that type of data on MIT backup media, especially on what should be magnetic tapes.

When I think about PC's at MIT in the 80s, I of course think of the author of PC/IP. So, do you have any idea how this MS-DOS data would appear on magtapes from MIT?

Thanks!

[larsbrinkhoff]

Classification results so far, on my limited corpus.

   2136 ITS DUMP
   1410 TOPS-20 DUMPER
    683 Unix dump
    261 Symbolics LMFS dump
    236 Unknown
    154 ITS DUMP (with label)
    106 Unix tar
    105 Error reading tape image
     78 VMS BACKUP
     15 TOPS-20 install
      8 FAT file system
      4 Unix cpio
      4 TOPS-10 BACKUP
      1 TOPS-10 FAILSAFE
      1 MIT/LMI dump

Breakdown of platforms:

Tapes	Percentage	Platform
5202	100	All
3720	72	PDP-10
793	15	Unix
341	7	Unknown/bad
262	5	Lispm
78	1	VMS
8	0.1	PC