[matroska-devel] Re: Comments needed on seeking part

Steve Lhomme steve.lhomme at free.fr
Fri Feb 14 17:31:44 CET 2003

En réponse à ChristianHJW <chris at wiesneronline.net>:

> you know i am not a coder, so this may be complete rubbish now, but look
> at
> it from a brainstoming point of view ;)  :
> Instead of having one big seek table at the end of the file, why not
> make it
> like that :
> 1. several seek tables at cluster 100, 200, 300, 400 ,etc.
> 2. Seek tabel at cluster 100 contains the entries for cluster 1 - 99 ,
> table
> at cluster 200 contains entries for 101 - 199, etc ....
> 3. To be able to cover different scenarios ( huge cluster sizes, etc. )
> you
> can define a value stating in the file header telling you after how
> many
> clusters a new seek table would come. In my scenario above this value
> was
> 100

As Cyrius said on IRC (yes, I'm a lurker) there's no use for a seek table or cue
entry if it's split in many parts throughout the file. If it is spread among the
file, it should always be the same one (reemission during a stream for example).

For the VDub case, I think the placeholder I just talked about should be enough.
Note that the Cue entries can still be at the end (should be much bigger than
the Meta Seek entry) and contain the position inside the Meta Seek entry. (cue
only define the position of a Cluster+Block according to a timecode, the Meta
Seek entries can be specified for any matroska element).

More information about the Matroska-devel mailing list