daniel/deepdrft
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 13 Wiki Activity

docs(plan): add Phase 1.1 Extended WAV format support

Browse Source 
Track the two EXTENSIBLE WAV sub-cases scoped out of the
WAVE_FORMAT_EXTENSIBLE PCM fix: non-PCM (IEEE Float) SubFormats and
padded 24-in-32 containers.
This commit is contained in:
daniel-c-harvey
2026-06-10 15:31:25 -04:00
parent d2a9475ba2
commit 38529a962a
1 changed files with 12 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
PLAN.md
+12
View File
@@ -23,6 +23,18 @@ What this means for the roadmap: the streaming substrate is solid. Future work c
These were flagged during the audit but classified as feature work, not defect fixes. They are listed in rough order of user-visible impact.
### 1.1 Extended WAV format support
- **What:** Two EXTENSIBLE WAV sub-cases that were explicitly scoped out of the `WAVE_FORMAT_EXTENSIBLE` PCM fix (which shipped support for `audioFormat=0xFFFE` with a PCM SubFormat — the Bandcamp WAV download case). Both are currently rejected at `AudioProcessor.ValidateAudioParameters` and fall back to default metadata. The inline comments at `AudioProcessor.cs` (SubFormat check ~L182–188, BlockAlign note ~L225–230) mark them as accepted gaps as of that fix.
- **EXTENSIBLE non-PCM SubFormats** — e.g. IEEE Float (32-bit float PCM, common in DAW exports). The SubFormat-GUID check accepts only PCM (`0x0001`) today; anything else is rejected outright.
- **Padded-container EXTENSIBLE** — 24-bit valid samples in a 32-bit container (`wValidBitsPerSample=24`, container `bitsPerSample=32`). The BlockAlign check fails because the valid-bit depth (24) doesn't match the container's block align.
- **Why it matters:** DAW exports — the dominant shape of source material as the collective uploads more of its own production — tend to be float WAV or padded 24-bit. The shipped fix covers consumer/Bandcamp WAVs but not the producer's working files.
- **Shape:** Both live in the same seam as the shipped fix (`AudioProcessor` validation + the `NormalizeToStandardPcm` storage step), but the work differs by case:
- *Float SubFormat:* requires float→integer sample conversion during the normalize-to-standard-PCM step (the vault stays integer-PCM so the streaming/decode pipeline is unchanged), or a Web Audio decode path that handles float directly. The conversion-at-storage option keeps the load-bearing streaming seam untouched and is the lower-risk path.
- *Padded 24-in-32:* relax `ValidateAudioParameters` to tolerate the BlockAlign mismatch when `IsExtensible`, then normalize to the valid-bit depth (24) during storage so the stored WAV is canonical.
- **Prerequisite:** None. Both are self-contained extensions of the WAV path that just landed; neither depends on the broader format-router work in 1.2.
- **Relationship to 1.2:** Distinct from it. 1.2 is new *containers* (MP3, FLAC, Ogg) behind a format router; this is additional *WAV variants* on the existing PCM path. If 1.2's router lands first, these become per-variant branches inside the WAV processor rather than new processors.
### 1.2 Audio format diversity
- **What:** Today `AudioProcessor`, `WavOffsetService`, and the JS decoder are PCM/WAV-only. `MimeTypeExtensions` already maps MP3, FLAC, Ogg, AAC, M4A — none are wired.