Eyevinn / moqlivemock / 25356779079

44%

Build Type

push

github

Committed by tobbee

Commit Message

feat(catalog): expose accurate per-packaging bitrate

Catalog Bitrate now reflects the actual wire footprint of each track,
correctly differentiating clear CMAF, encrypted CMAF (cenc/cbcs), and
LOC. Previously all CMAF tracks shared one estimator (constants from
draft-09 mp4ff layout) and LOC reported the raw sample bitrate — so
encryption overhead and LOC framing were invisible to clients.

- calcCmafBitrate measures actual chunk size via GenCMAFChunk for the
  track's real configuration (sample batch, encryption scheme,
  subsample density). Self-consistent under future packaging tweaks.
- A per-object MoQ overhead constant (cmafObjectOverheadBytes = 8) is
  added on top of the measured chunk bytes to model the wire framing
  paid per object: ObjectID + payload-length + extension-count + status
  varints per draft-ietf-moq-transport-16. Subgroup header amortises
  small enough to round into this constant.
- calcLOCBitrate adds the same MoQ object overhead plus a 9-byte LOC
  Timestamp extension property per object (draft-ietf-moq-loc-02
  §2.3.1.1), and (for video) the VPS/SPS/PPS prepended to every IRAP
  frame.
- GenCMAFChunk enables mp4ff OptimizeTrun, which promotes constant
  per-sample fields into tfhd defaults. Audio per-extra-sample drops
  from 16 B to ~3 B (4 B sample-size, 0 B for tracks with constant
  sample size like Opus / AC-3), shaving real bytes off every audio
  chunk on the wire.
- Default -audiobatch goes from 2 to 1 so audio chunking matches LOC
  (one frame per object). The wire-cost gap between packagings is now
  visible directly: AAC LOC ~+4 %, CMAF clear ~+33 %, CMAF cenc ~+54 %.

Coverage Stats

35 of 48 new or added lines in 2 files covered. (72.92%)

61 existing lines in 3 files now uncovered.

1820 of 4130 relevant lines covered (44.07%)

1070.14 hits per line