FLOSS Multimedia Support in Fedora

Gregory Maxwell gmaxwell at gmail.com
Mon Feb 9 00:45:36 UTC 2009


2009/2/8 Martin Sourada <martin.sourada at gmail.com>:
> Hi,
> sorry for cross-posting but I wanted all interested parties to not miss
> this email ;-)
>
> There has been some media related discussion in the -devel list and one
> of the points I have taken form it is that we should promote FLOSS
> multimedia and don't blame others for doing it (even if it's done in a
> not ideal way)... Now, the problem is that the actual support of these
> in our system is crappy.

I don't think that is the correct take-away.  Everything works in gstreamer.

The correct conclusion is that support for free formats in media
infrastructure primarily designed, built, and maintained for non-free
formats currently stinks for the free stuff.

It's another indicator that Fedora should discontinue shipping these
non-free media focused infrastructures rather than continuing to patch
the non-free stuff out.

Of course, support for the free formats should be fixed in that, but
that really shouldn't be a Fedora issue.

[snip]
> Also
> feel free to reference upstream bugs about the mentioned issues. The
> videos used for reference testing are available at my fedora people page
> [2].

It's probably important to have a more complete test.

Right now you're using what is probably the simplest possible test:
Just a regular file with video in it. That case is important, but it
is also important to test more complicated cases, such as tests
including audio, subtitles, oddball muxing, and chaining.   It's no
good if a player works for your simple case but crashes as soon as
someone gives it a subtitled input.

It would also be good to try all mixtures of (FLAC|Speex|Vorbis) *
(Theora|Dirac) * (w/ | w/o subtitles) * (OGG|Mkv) which is 24
combinations by itself.
(although I don't know how worthwhile the Mkv testing is— as far as I
can tell it's mostly unseen outside of the movie piracy / anime fansub
sites; and at those its usually used with H264)

For Ogg it's also important to test seeking, because some applications
have historically gotten it wrong. Seeking with Ogg should be done
with Newton–Raphson (ideally) or bisection. There have been broken
applications (ffmpeg in the past; for example) which seek in Ogg with
a linear scan starting at the file beginning, which is quite
intolerable for any reasonably large file.  Unfortunately this test
needs a large (~hundreds of MBytes) test file, so it's hard to
distribute a test.


Some of the proprietary-codecs focused tools provide their own home
grown implementations of the codecs (i.e. ffmpeg). These often do not
implement the full spec, so its important to test their behaviour.

Here are some resources for doing that:

Some good tests for Ogg/Theora:
http://v2v.cc/~j/theora_testsuite/

There are test vectors for Vorbis here:
http://people.xiph.org/~xiphmont/test-vectors/vorbis/




More information about the Fedora-art-list mailing list