[Cellar] Clarification on blocks with a negative timecode

[Michael Bradshaw <[e-mail]>]

Blocks can have a negative timecode in the following situations:


   - Cluster Timecode + Block timecode < 0. This may happen if the
   Cluster's Timecode is near zero and the Block's timecode is sufficiently
   negative.
   - CodecDelay > Cluster Timecode + Block timecode. I have a WebM file
   created by ffmpeg with an Opus audio track that has a CodecDelay of 6.5ms,
   and since the first block's timecode is zero, its computed timecode is
   -6.5ms. mkvinfo shows this frame has having a timecode of 0.000s, but
   ffprobe shows a PTS of -0.007ms (rounding up).


The notes say that "Blocks with a negative Raw Timecode are not valid."[1]
This page, however, does not mention CodecDelay's role in computing
timecodes.

Questions:


   - Should the Specification Notes talk about CodecDelay's role in
   computing a block's timecode?
   - Assuming the answer to the previous questions is "yes," is a block
   with a negative timecode due to a CodecDelay still invalid (i.e. is
   CodecDelay a part of Raw Timecodes)?


--Michael

[1]: http://matroska.org/technical/specs/notes.html