[MPlayer-dev-eng] Proposed support for Kate streams

[Evgeniy Stepanov]

On Tuesday 18 March 2008 00:32:13 Reimar Döffinger wrote:
> On Mon, Mar 17, 2008 at 10:20:03PM +0100, Reimar Döffinger wrote:
> > On Mon, Mar 17, 2008 at 11:04:57PM +0200, Ivan Kalvachev wrote:
> > > On Mon, Mar 17, 2008 at 10:00 PM, Reimar Döffinger
> > >
> > > <Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> > > > On Mon, Mar 17, 2008 at 06:44:26PM +0000, ogg.k.ogg.k at googlemail.com 
wrote:
> > > >  > I've patched MPlayer to support Kate streams in Ogg and Matroska
> > > >  > containers. Kate is a text codec (for subtitles, etc). More
> > > >  > information can be found there:
> > > >  > http://wiki.xiph.org/index.php/OggKate
> > > >  >
> > > >  > Note that the support is just as simple text subtitles (eg, none
> > > >  > of the motions, regions, styles are supported (yet)).
> > > >  >
> > > >  > For reference, here's the current patch, but against a svn version
> > > >  > a few days old already.
> > > >  >
> > > >  > Would you be interested in this ?
> > > >
> > > >  Sorry for being offensive, but I would prefer to wait and hope it
> > > > dies out before it ever gets used much, since it extremely smells
> > > > like another case of NIH-syndrome caused by too much use of Ogg. Esp.
> > > > considering this at the wiki:
> > > >  "I found other formats, such as USF (designed for inclusion in
> > > > Matroska) and various subtitle formats, but none were designed for
> > > > embedding inside an Ogg container."
> > > >  Which in my ears screams: "Lets reinvent the wheel instead of
> > > >  fixing/improving Ogg container".
> > >
> > > Does this mean that the main design goal is these subtitles to be
> > > almost unusable without ogg?
> >
> > I think the design not so much, the implementation in MPlayer is more of
> > a problem though, since the "decoding" is done in the demuxers (support
> > for ogg and mkv is added), instead of in mp_common.h.
> > Probably because demux_ogg still uses the old and broken way of setting
> > subtitles (modifying vo_sub directly instead of creating a proper
> > subtitle stream).
>
> And I also forgot: this means that your patch might be much less helpful
> than you think, since I intend to make the libavformat demuxer the
> default over our one rather sooner than later, and I am rather convinced
> that the FFmpeg developers will not accept a demuxer using an external
> library like libkate (or what you call it, btw. I think the naming is quite
> unfortunate since it collides with the name of a KDE editor, luckily so far
> they have only a libkatepart.so, still seems risky to me though). Which
> means it would have to be redone to move the decoding in
> mpcommon.c or it would only work with an explicit "-demuxer ogg" then.

I think it won't hurt to apply the ogg demuxer part of the patch. Won't make 
it any worse, it is an unmaintainable mess already. But as for other 
demuxers, I fully agree with you, it's unacceptable. Subtitles should be 
exported as binary packets with type='k' and decoded in mpcommon.h.

Sure, I don't like the idea of one more 'just for ogg' subtitle format, but 
that should no be a reason to reject a clean, nonintrusive patch, if one is 
proposed.

Slightly off-topic, it seems too early to switch to lavf demuxer at least for 
ogg and avi: the former does not support seeking to keyframes, and the latter 
takes too much time to open because it reads full odml index. On the other 
hand, matroska is almost ready for the switch, with only missing feature 
being the support of 'default' track attribute, which is not really important 
and could be easily implemented anytime.