[pulseaudio-discuss] rtp question

Tanu Kaskinen tanuk at iki.fi
Wed Feb 20 10:06:31 PST 2008

On Wed, Feb 20, 2008 at 09:24:22AM -0800, Erich Boleyn wrote:
> A bigger question is, then:  What is the whole point of RTP?  If you are
> multicasting to multiple receivers without resampling, then there is
> a bufferring problem which will guarantee drop-outs or gaps.

I'd guess the occasional glitch isn't so annoying as to
render RTP completely useless.

> Is this in the plans to be fixed?  (or am I basically volunteering... ;-)

I have already pasted this link earlier:

Last time, when I still thought module-rtp-recv does
adaptive resampling, I interpreted Lennart's comment to mean
that rtp would become as good solution as module-combine +
module-tunnel-sinks (in syncing sense). Now I think it means
that the clock drift elimination will eventually be copied
from module-combine to module-rtp-recv - the latency
difference problem would remain unsolved, but that may be a
minor problem anyway.

> FWIW, I noticed the other thread "Controlling where module-rtp-send
> sends   multicast packets?" which is talking about the use of
> module-combine and tunnelling, but I have a many-to-many situation
> and without a central "owner" of audio it seems a bit weird to do it
> the other way.

I think that solution can be used in your case as well,
hacking together a centralized control thingie would just be
somewhat trickier, I believe.

Every sending machine would have as many tunnel-sinks as
there are receiving machines. There would be one
module-combine instance per stream (I guess that in your
case this means one per sending machine).

Tanu Kaskinen

More information about the pulseaudio-discuss mailing list