[Asterisk-Dev] Silence suppression in asterisk/chan_sip?

[Paul Cadach]

Kevin P. Fleming wrote:
> I've thought about this too, but I think the overhead of every RTP
> stream using the timer separately would just be too much. I think the
> best solution is for rtp.c to create a thread when it loads, use the
> zaptel timer to clock that thread, and have all outbound RTP sent by
> that thread based on a reliable clock source.

This thread should handle self-generated packets (comes from Zaptel, generators,
etc., i.e. non-VoIP sourced). At least it could help a little with timestamp
generation/calculation.

> This would also help with RTP bridging, since there would only be one
> bridging thread, not one for each call.

Bridged calls should use its own timing, while mixed calls (conferences, etc.)
and generators (and MOH, of course) could use internal Asterisk's timer (based
on zaptel, RTC or system clock source). I had done someting like that for 0.5.0,
just for generators (including MOH) - all worked fine until I moved to newer
version.

Also, IMHO there should have different SSRC for each voice stream (Asterisk's
internal streams is one SSRC because uses its own timing and any other
enpoint(s) have different SSRC(s)).. RTP's 'M'ark flag could notify remote party
about clock source is changed.


WBR,
Paul.