Advanced Dungeons & Dragons — Slayer, 3DO Interactive Multiplayer, USA. One pressed CD-ROM, documented from a CHD without running or emulating anything on it.
A dungeon crawler that spends two thirds of its disc on a soundtrack it did not compress, whose operating system is byte-identical with the disc next door on 115 files of 116, and which names nobody in any string — but says who made it, four times, in pixels.
| container | one CHD v5, 200,769,302 B, SHA-1 126b1dcc03007b1482cd9583edb2c2754cb5effa |
| pressing | one MODE1_RAW track, 151,340 sectors, 45.4474 % of a 74-minute CD |
| audio tracks | zero — the first of four 3DO discs here with none |
| pressed | 1994-08-16 09:29:56, from the disc's own /rom_tags |
| file system | Opera, not ISO 9660. 582 files, 24 directories, 272,214,595 B |
| hashes | 573 distinct of 582, seven duplicate groups |
| mastering fill | iamaduck, 17,729 sectors = 11.7147 %, in two regions |
| unowned sectors | 0; double-claimed 0; 300 zero sectors past the declared volume |
| the thesis | 67.4093 % of the user area is recorded sound — 81 AIFF files, codec NONE on 81 of 81, 22:30.04 |
| identified | 87.6184 % of the user area, the highest in the collection — and the least impressive, because the object did the work |
| the game itself | one 299 KB ARM binary, 370 cel containers, 2,516 frames, a font: under 4 % |
-
It has a twin, and the twin is 68 bytes away. Its
/Systemand Super Street Fighter II Turbo's are 115 identical files of 116. The one difference is a serial-port driver, and the difference is akprintfcall and its format string — five ARM instructions and 44 bytes of"Stamped event pod %d position %d generic %d". 08 -
The container nobody had walked, walks.
ANIM/CCB: 370 files closing at residue zero, 370 of 370, 3,782 chunks, onePDATper frame and 2,516 of 2,516 frames rendered. The reader for the chunk rule had existed for two discs and had never been aimed here. 06 -
The disc carries its own pressing date, in a 32-bit field of
/rom_tags, and it reorders the collection: Slayer was pressed 147 days before the disc it shares an SDK build with. The platform notes had already found this and asked a fourth disc for exactly this point — which this session did not read until after it had re-derived it. 11 -
Two hundred and nine megabytes of uncompressed music, when the platform's own 2:1 codec was on the same disc and the twin used it. SDX2 would have freed 102,379,300 bytes. The disc was 45 % full, so it cost nothing — and the object still supplies evidence about why. 09
-
The credit roll is 17 still screens, 20 people, and this repository publishes none of their names — because the owner of the disc was asked and chose that for this object, not because anything forbids it. The standing line is that what is on the disc is published and only reach is withheld; the clause permits the names and so does the house rule. 10
| 00 | predictions, written before anything was opened — 36 clauses, two columns |
| 01 | what it is, on which denominator, and what it keeps and throws away |
| 02 | the sheet: every figure with the command that remakes it |
| 03 | Opera a fourth time, and a label with two lengths that are both right |
| 04 | all 151,340 sectors, and 2,928 zero sectors of which only 300 are empty |
| 05 | the console does not read the filename, and that is why nobody noticed it was wrong |
| 06 | the ANIM container: 370 files close, 2,516 frames render |
| 07 | 41 ARM images, and the two that break the rule are the game's |
| 08 | the smallest useful diff this collection has: one debug print |
| 09 | the thesis, the SDX2 arithmetic, and 43 creatures with 36 voices |
| 10 | three films, a credit roll counted not transcribed, and the decision |
| 11 | a pressing date in 32 bits, and a question the third disc left open |
| 12 | 114 crossings of 573, all of them 3DO, and the twin |
| 13 | what was written into the platform checklist, and what it already knew |
| 14 | the refusals, the negative controls, and two tools that cannot say no |
| 15 | every wrong figure, left visible: the brief's, the pre-briefing's, this session's |
| 16 | 18.45 of 30 open against 19.50 predicted |
The measured listings, so every table above can be checked without the disc:
listing.txt (582 entries), sha1-all.txt (582 records), sectors.txt,
chd.txt, label.txt, containers.txt, frames.txt, romtags.txt,
aif.txt, reuse.txt, account.txt, crossall.txt, sdkdiff-ssf2t.txt,
and owner-observations.txt — what the person who owns the disc reported,
attributed, including the one check that ran from bytes to a person and back.
446 Python files. Three were written here and one was rewritten:
animwalk.py— walks theANIM/CCBcontainer, censuses its chunks, and renders everyPDATas a frame. Seven negative controls, one positive;romtags.py— reads/rom_tags, checks the0x02record against a directory listing, and argues the epoch with a table instead of a claim;aiffreuse.py— finds re-used sound that a hash census cannot see: exact payloads, contained payloads, and same-length correlations, with an envelope ranking that is explicitly not a test;account3do.py— rewritten for this disc, third session running, and the two changes are named in its docstring: the AIFF bucket is labelled with the codec the container declares rather than with a neighbour's, and the cel container is walked rather than signature-scanned.
Nothing else was modified. Twelve inherited tool defects are reported in 15 and none is fixed, because fixing an inherited tool silently is worse than reporting it.
No byte of the product, in any form it takes: no audio, no cel, no film, no
frame decoded into another format, and no name from the credit roll. The
.chd, _work/ and prompt.txt are in .gitignore, verified in both
directions with git add -An and git ls-files.