[Matroska-devel] Re: Dirac Video Codec

Steve Lhomme steve.lhomme at free.fr
Tue Oct 5 12:58:13 CEST 2004


Arioch /BDV/ a écrit :

> The stars so gaily glistened... (Tue, 05 Oct 2004 09:28:18 +0200 @352)
> ...while the fading voice of Steve whispered through the darkness:
> 
>  SL> To be more precise, the frames have a relative timecode to the Cluster.
>  SL> The Cluster has an absolute timecode.
> 
> I wonder, if for long files, there was sense in 3 levels of hierarchy.
> Say,  >>SuperCluster{timecode reltive to file start = absolute}[ ...SubClusters{timecode relative to SuperCluster}(Blocks)]<<
> 
> IF such a Supercluster, for example, would have it's own seek table, etc, that maybe could be better for seeking remote(streaming) files.
> Also for long files (real long) absolute timecode values might become too long?

Actually what you call SuperCluster is the Segment. Out of the segment, 
the "absolute" timecode doesn't have a real meaning. At least the way 
it's done now.

Because the Segment *has* a DateUTC that is supposed to give the 
_absolute_ date when the content was created (the reference being the 
beggining of this millenium).

Also with file linking the timecode of each linked Segment is contiguous 
to the one of the previous. But we could imagine a global delay at the 
Segment level to have the timecode for each Segment start at 0 and use 
the delay during linking.

We also didn't state yet if a Chapter section can cover more than the 
Segment it's contained in. The current usage would be : no.



More information about the Matroska-devel mailing list