> > > Of course, you can offload some parts of the decoding with
> > > XVmc, but that functionality is not widely spread and doesn't
> > > work for all codecs.
> > 
> > This brings up the question of how much bandwidth do you need between
> > the client and server if the client decodes?  E.g. is Ethernet fast enough?
> > If not, then we have to decode inside the X-video-server box.
> 
> for 720p, subsampled to 4:2:0, 30fps:
> 1280*720*2*30 = 53MByte/s

30 fps, not 60?  Hopefully the fans of fast action will not be too mad over
losing half the frames.

So 1080i would be 1920*1080*2*30 = 124416000 = 125 MB/s = 119 MiB/s  ?
124416000 * 8 = 995328000 bps
That would be *very* difficult, probably impossible, to squeeze into a
1 Gbps pipe when you consider protocol overhead and such.

> But actualy, you do not want to have video and audio output
> on different machines connected over a slow network. The
> delay becomes unpredictable and thus A-V synchronisation
> is very bad.

The X-server-box would have audio output.

> > > Modern schemes that
> > > are wavelet based exceede this by a factor of 10 or even more.
> > 
> > And I suppose we can expect a new incompatible-wrench-in-the-gears
> > every couple of years for quite awhile.  So we'll have to make a
> > list of what the X-video-server will decode, and anything incompatible
> > will have to be transcoded on a general purpose computer.  Which is a
> > major PITA.
> 
> Expect to do that already now. MPlayer currently supports something
> in the range of 200 different video codecs. libavcodec (the main
> decoding library) commes currently with over 50 codec schemes, most of 
> them decoding more than one video codec.
> Even if you want to limit yourself to the most used codecs you'll
> get way more than you can support (i'd say something in the range
> of 20 codecs).

200?  Obviously it is far too easy to create your own codec.

So let's compile a *short* list of the most useful codecs.

mpeg 1/2/4 covers OTA and DVD and more
IIRC Blu-Ray and HD-DVD use H.264 ?
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to