[gst-devel] Re: RFC: GNOME 2.0 Multimedia strategy

[Christian Schaller]

Hi people,
I thought I should try to summarize the discussion on IRC last night in
order to let everyone stay informed. I might be misquoting people here,
but I guess if I am people will correct me :)

The discussion had mainly four participants being Elliot Lee, Sander
Vesik, Erik Walthinsen (lead GStreamer developer) and myself. Have to
admit that I probably the least active one in the discussion leaving the
technical bits to Erik.

For those not reading my previous mail the GStreamer homepage is
(http://www.gstreamer.net)

Anyway both Elliot and Sander where sceptical about the use of asm
cothreads in GStreamer. 
The following decisions/conlusions was made.
a) The use of cothreads is easily replaceable and Erik plans on making a
pthreads fallback for it. The pthreads version Erik will start hacking
on in about 2-3 weeks.

b) In order to prove one way or the other that the use of cothreads
doesn't pose a big risk we will try to stage some kind of largescale
user testing in order to confirm wether cothreads in fact works on
99% of the systems it claims to support.

c) Sander was sceptical of the performance gains cothreds gave so he and
Erik will conduct some performance tests comparing cothreads and a
pthreads implementation. 

Think that was the core of that discussion.

Ok, the other part was concerning soundservers. Elliot objected to the
use of GStreamer for `gnome_play_sound("bleep.wav")` stuff feeling it
was overkill and felt that this type of sounds should go directly to the
soundserver.  He did however agree that we should use GStreamer for the
applications currently in gnome-multimedia.

Personally I think we should use GStreamer for both, since when he have
GStreamer there (which we agreed upon we should) I think we should use
it for everything, especially since that would give us full soundserver
independence. A soundserver might however be a easier dependance for
gnome-libs, since GStreamer at least if not packaged wisely could pull
in a  lot of sub-dependancies. (See GStreamer roadmap for details:
http://www.linuxrising.com/files/gstreroadmap2.html)

There where a lot of discussion concerning the advantages of different
solutions like esd, asrtsd, asd and gnostream and using GStreamer for
also system 'dings' would make us much more flexible and even allow
users & distributions to choose for themselves what soundserver to use.

This is however not a major point and Elliot did say he would check in
code to gnome-libs to make switching soundserver easier.

Elliot would also do some research on media/soundserver technologies in
order to send a recomadation to this this list for what such technology
to use in conjunction with GStreamer under GNOME 2.0.

Conclusion was that I think people agree on GStreamer being part of
GNOME 2.0 with the only objections being on implementation details
and practical use. Issues which easily can be solved.

Christian