[MPlayer-dev-eng] NUT, what's left

Luca Barbato lu_zero at gentoo.org
Sat Sep 17 14:45:22 CEST 2005


Michael Niedermayer wrote:

> 
>>they 
>>   should all be marked as key frames? (doesn't really matter).
> 
> 
> dunno
> 

If they are there is no need for info_startcode.

> 
>>3. why do we still need info_startcode? it can be removed.
> 
> 
> yes, but maybe its usefull for error recovery, it adds no complexity
> to the muxer at all and negligible complexity to the demuxer IMHO
> 

info_startcode could be the same as frame_startcode

> 
>>4. what do you mean by pts difference of zero? it's identical to the 
>>   stream's "last_pts"? why not just use '0'? it makes more sense (they are 
>>   at the begginning, apply globally, sound like '0' to me).
> 
> 
> no, there might be info frames with non zero pts, only the ones
> with global meaning must be at the start and have pts=0
> 

basically if stream_id == infodata && pts = 0 you have global and
repeatible info, that means that you make the demuxer a bit more complex
since that is an exception to the muxing/demuxing rules.

I'm not sure if would be more problematic having pts = 0 frames here and
there or info packets.

Probably in the end would be useful using the info_startcode to mark the
pts=0 info frames and frame_startcode for the pts!=0.

So sum up:

merge 2 startcodes or use them that way to avoid conditionals in the
demuxing code?

> 
>>5. do these streams deserve an index?...
> 
> 
> why not
> 

The stream index for info stream would be quite useful, it should ignore
the subsequent pts=0 or not?

lu

-- 

Luca Barbato

Gentoo/linux Developer		Gentoo/PPC Operational Leader
http://dev.gentoo.org/~lu_zero




More information about the MPlayer-dev-eng mailing list