Skip to content

Docs: re-measure the collection, and correct a partition count by one - #29

Merged
bmxcode merged 1 commit into
mainfrom
fix/refresh-tested-against
Aug 21, 2026
Merged

Docs: re-measure the collection, and correct a partition count by one#29
bmxcode merged 1 commit into
mainfrom
fix/refresh-tested-against

Conversation

@bmxcode

@bmxcode bmxcode commented Aug 21, 2026

Copy link
Copy Markdown
Owner

The README's "Tested against" block still described the run before D15
read every AKAI partition, and disagreed with docs/README.md on the same
page. Everything in it is re-measured over all 79 images.

discs converted 69 of 79 -> 72 of 79
samples 47742 -> 89125
stereo pairs 4422 -> 15796
damage skips 20 -> 97
time 31s -> 46s
distinct rates 908 -> 1047 (6000-49999, was 6000-48000)
silent whole-file 10 -> 275

The byte-identity claim was quoted twice on one page at two different
sizes -- 22 320 payloads in one paragraph and 40 244 in another. It is
now one statement: 70 of 70 discs match exactly, 89 125 payloads, zero
mismatches, comparing multisets of SHA-256 over the PCM per disc. The
two audio CDs are excluded and said to be excluded, because a track is
cut from a stream by a cue and is not a run of bytes anywhere on the
disc.

The silence count moving from 10 to 275 is not a regression. Every one
of the 275 matches the disc exactly, so it is what the discs hold: 267
are the blank 15G-KIT...Z slots on ProSamples vol.15, six are the Proteus
"Dead Air" samples the page already described, two are on a Roland disc.
D15 unlocked 3.4x more AKAI content and the blank kit slots came with it.

Also corrected: docs/formats/akai-fs.md and three comments in
fs/akai.py said the 44 AKAI discs hold 276 partitions. The table
declares 384 and 275 are present in the images. Measured twice by
separate routes, and 275 is what len(list(partitions(...))) sums to --
the same expression test_akai_discs_list_their_volumes_and_files already
pins per disc for nine of them, which is what makes this a correction
rather than a preference. The figures beside it were re-checked and are
exact: 2 154 volumes, 68 997 files. The earlier 276 could not be
reproduced.

ADR-0023 also says 276 and is left alone, as ADR-0019 was left alone
through the disc renames: an ADR records what was measured when the
decision was made. The correction is in the changelog instead.

No source behaviour changes. The only non-doc edits are three comments.

🤖 Generated with Claude Code

The README's "Tested against" block still described the run before D15
read every AKAI partition, and disagreed with docs/README.md on the same
page. Everything in it is re-measured over all 79 images.

  discs converted   69 of 79  ->  72 of 79
  samples               47742  ->  89125
  stereo pairs           4422  ->  15796
  damage skips             20  ->  97
  time                    31s  ->  46s
  distinct rates          908  ->  1047  (6000-49999, was 6000-48000)
  silent whole-file        10  ->  275

The byte-identity claim was quoted twice on one page at two different
sizes -- 22 320 payloads in one paragraph and 40 244 in another. It is
now one statement: 70 of 70 discs match exactly, 89 125 payloads, zero
mismatches, comparing multisets of SHA-256 over the PCM per disc. The
two audio CDs are excluded and said to be excluded, because a track is
cut from a stream by a cue and is not a run of bytes anywhere on the
disc.

The silence count moving from 10 to 275 is not a regression. Every one
of the 275 matches the disc exactly, so it is what the discs hold: 267
are the blank 15G-KIT...Z slots on ProSamples vol.15, six are the Proteus
"Dead Air" samples the page already described, two are on a Roland disc.
D15 unlocked 3.4x more AKAI content and the blank kit slots came with it.

Also corrected: docs/formats/akai-fs.md and three comments in
fs/akai.py said the 44 AKAI discs hold 276 partitions. The table
declares 384 and 275 are present in the images. Measured twice by
separate routes, and 275 is what len(list(partitions(...))) sums to --
the same expression test_akai_discs_list_their_volumes_and_files already
pins per disc for nine of them, which is what makes this a correction
rather than a preference. The figures beside it were re-checked and are
exact: 2 154 volumes, 68 997 files. The earlier 276 could not be
reproduced.

ADR-0023 also says 276 and is left alone, as ADR-0019 was left alone
through the disc renames: an ADR records what was measured when the
decision was made. The correction is in the changelog instead.

No source behaviour changes. The only non-doc edits are three comments.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@bmxcode
bmxcode merged commit bd27ce2 into main Aug 21, 2026
1 check passed
@bmxcode
bmxcode deleted the fix/refresh-tested-against branch August 21, 2026 08:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant