[MPlayer-dev-eng] mplayer-ww (mplayer fork!)

[Reimar Döffinger]

On Thu, Nov 04, 2010 at 08:28:03PM +0100, Reimar Döffinger wrote:
> On Thu, Nov 04, 2010 at 01:06:51PM +0100, Diego Biurrun wrote:
> > On Fri, Oct 29, 2010 at 07:19:01PM +0200, Reimar Döffinger wrote:
> > > On Fri, Oct 29, 2010 at 09:52:35AM -0400, compn wrote:
> > > > william wangs' pmp demuxer got somewhat ignored when he posted it,
> > > > but since then he has developed mplayer-ww for windows. applying
> > > > sherpya's patches (dsnative, winamp dsp filters). but also extending it
> > > > to support directshow demuxers. also some fixes to opengl/directx
> > > > in w7/vista.
> > > > 
> > > > http://sourceforge.net/projects/mplayer-ww/
> > > > http://www.mplayer-ww.com/eng/changelog.html
> > > > 
> > > > his mplayer can play the pmp samples:
> > > > http://samples.mplayerhq.hu/playstation/psp/
> > > > 
> > > > any plans to merge some of this stuff?
> > > > the pmp demuxer could be a good first step.
> > > 
> > > I don't really intend to accept any new demuxers for MPlayer.
> > 
> > I disagree here.  Our demuxers are unlikely to ever get removed, since
> > they are a useful alternative to libavformat.  Plus, the PMP demuxer is
> > much more likely to get ported from libmpdemux to libavformat than from
> > some obscure MPlayer fork.
> 
> I don't think there is any significant likelyhood, no demuxer or decoder
> was ever ported to FFmpeg, they were at best "reverse engineered" (in
> case of RTjpeg that's not even an exaggeration) and reimplemented.
> 
> > So if somebody cleans up the PMP demuxer for inclusion, I'd say we
> > should commit it.
> 
> That would mean spending developer time and review time to add it to
> MPlayer, then spending more developer and review time to port it
> to FFmpeg and the we'd still have only a GPL-only demuxer.
> Having someone derive a specification from the code and then (same or
> other person) implementing it for lavf is a much more sane way to do it
> unless I vastly underestimate the complexity of the PMP format.

I don't think I did, see here:
http://wiki.multimedia.cx/index.php?title=PMP
Even if I gave up my resistance, the demux_pmp code would have to
be completely rewritten to be acceptable, e.g. it stream_reads
directly into C structs (alignment and endian issues), does
something like 6 seeks frames per frame (none of which are
necessary) and comes with progressively worsening performance
since to demux a frame it adds up the sizes of all previous
ones - again and again.