2009/5/6 Vladimir Pantelic <pan_at_nt.tu-darmstadt.de>:
> Alex Bennee wrote:
>> That would be pathalogical. At the moment I've a fixed size array in
>> the codec that will deal with up to 10 mdat atoms in a file.
< snip>
>
> do you really care about the mdat atoms? The offset data for the samples is
> not mdat relative, so you don't need to keep the mdat info in memory, do
> you?

To be honest I'm not sure. I'm kinda learning the MP4 container format
on the fly as I do this. If all the seek information is already
contained in the moov section and the offsets are relative to the
whole file then yes, I can just ensure we parse the full file before
playback and seek back to the start of the first mdat when we start
playing.

The real test will be having some good test files that people can't
currently play and using that prove we have covered all the corner
cases.

-- 
Alex, homepage: http://www.bennee.com/~alex/
CV: http://www.bennee.com/~alex/cv.php