I suspect this is one of those questions where you know the answer ;-) Yes,
you're of course right. You inject a direction=active in the failure_route,
so that the GW will wait for the media before sending. However, I have had
some issues with some clients that I haven't really figured completely out
yet and I'm trying to be conservative. Do you have experiences where this
work? If so, which GWs and clients do you have good experiences with?
When we are first at it, there is another advanced case also:
Where one SIP client is non-NATed and supports connection-oriented media,
you don't have to use rtp proxy. The challenge is to know that a) the
client is really public and does not appear to be public due to STUN usage
b) it supports connection-oriented media (Honestly, I don't really know
which clients do, AFAIK only GWs have implemented this so far, but there are
probably exceptions)
So, if you don't use STUN at all and know your clients well, you should be
able to reduce the use of rtp proxy even more...
g-)
Juha Heinanen wrote:
Greger V. Teigre writes:
- PSTN-SIP when the SIP client is behind
symmetric
greger,
why does pstn-sip require use of media/rtp proxy if sip client is
behind symmetric nat? can't media/rtp proxy send to the same ip/port
where media is coming from the client?
-- juha