[FFmpeg-devel] [RFC] comment mpegvideo_split

Måns Rullgård mans
Fri Apr 4 10:01:01 CEST 2008


Michael Niedermayer <michaelni at gmx.at> writes:

> On Fri, Apr 04, 2008 at 02:18:16AM +0100, M?ns Rullg?rd wrote:
>> Michael Niedermayer <michaelni at gmx.at> writes:
>> 
>> > On Thu, Apr 03, 2008 at 11:20:15PM +0100, M?ns Rullg?rd wrote:
>> >> Michael Niedermayer <michaelni at gmx.at> writes:
>> >> 
>> >> > On Thu, Apr 03, 2008 at 11:05:18PM +0200, Baptiste Coudurier wrote:
>> >> >> Michael Niedermayer wrote:
>> >> >> > On Thu, Apr 03, 2008 at 03:52:19PM +0200, Baptiste Coudurier wrote:
>> >> >> > [...]
>> >> >> >> FYI:
>> >> >> >> ffmpeg -i <file.m2v> -vcodec copy test.nut does not work as is.
>> >> >> > 
>> >> >> > fixed
>> >> >> > 
>> >> >> 
>> >> >> Hehe yes, though I doubt mplayer using libmpeg2 can decode mpeg2
>> >> >> in nut :/
>> >> >
>> >> > It likely would just need a 2 line change ...
>> >> >
>> >> > Anyway, maybe we should just disable the mpeg1/2 header split
>> >> > thing. Ive read the nut spec again and it only recommands to put
>> >> > global headers in
>> >> 
>> >> MPEG2 does not have global headers.  Splitting the stream like this is
>> >> arguably creating an invalid MPEG2 stream.  If Nut wants to store it
>> >> in a special way, the Nut demuxer should reassemble it to a compliant
>> >> form.
>> >
>> > If MPEG2 does not have global headers then nut does not say anything
>> > about spliting them.
>> 
>> Then why was something acting as though that were the case?
>
> Because i thought that mpeg1/2 had been well designed and the global
> parts seperated.

MPEG2 was primarily designed for streaming (TV), where delivery of a
global header is tricky.  I doubt it was even considered.

What constitutes a good designed, depends to a large extent on the
intended application.

> But the sequence header can change thus it does not qualify as
> global. Though i must admit that calling it "sequence header" is
> slightly wrong, as the quant tables it contains do not apply to the
> whole sequence.

It applies until the next sequence header, but I do see your point.

-- 
M?ns Rullg?rd
mans at mansr.com




More information about the ffmpeg-devel mailing list