Hi Andreas,
o Andreas Granig on 01/14/2011 02:19 PM:
Hi Stefan,
On 01/14/2011 01:26 AM, Stefan Sayer wrote:
In passive mode, the remote RTP address:port is learned by the first
received packet. After the first packet is received, passive mode is
switched off.
Does that mean that SEMS won't send any packets to a side until one is
received from it? If so, that will probably break scenarios where T.38
is used from the beginning. We've seen some implementations where a T.38
endpoint won't send any packet until it receives at least one.
The term 'passive' mode is indeed misleading - it actually sends
packets to the address announced in SDP, just only when it receives a
packet on the local port, the remote address is updated. I'll update
the docs.
Actually, I think that in that old NAT traversal draft which proposed
this comedia-style symmetric RTP with direction 'active', it was
proposed to not send RTP from the 'server' until a packet was received
from the 'client'. But I guess that in reality this creates more
problems, like you mentioned.
We solved this (in our rtp proxy) by sending packets to the address
announced in the SDP, then updating that information once a packet is
received from that side. It still won't work with such T.38 clients
behind NAT, but it'll do its job in some scenarios.
Yes, that's the was how it is implemented now in sbc_rtp_relay.
Stefan
Andreas
------------------------------------------------------------------------
_______________________________________________
Semsdev mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/semsdev
--
Stefan Sayer
VoIP Services Consulting and Development
Warschauer Str. 24
10243 Berlin
tel:+491621366449
sip:[email protected]
email/xmpp:[email protected]
_______________________________________________
Semsdev mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/semsdev
