[OpenSIPS-Users] timeout on "100 Trying"

[Iñaki Baz Castillo]

2010/8/31 Julien Chavanton <jc at atlastelecom.com>:
> Hi, we have a core switch that can reroute calls on timeout, it is using
> timeout on "100 Trying"
>
> I was thinking opensips could wait for "100 Trying" comming from the remote
> switch :
>
>
> UA1 -INVITE-> P1
>               P1 -INVITE-> UA2
>               P1 <--100--- UA2
> UA1 <--100--- P1
>
>
> From t_relay doc
> 0x01 - do not generate an 100 trying provisional reply when building the
> transaction. By default one is generated. Useful if you already pushed an
> stateless 100 reply from script.
>
> Will Opensips forward 100 Trying ?

Never. 100 response is always hop-by-hop rather than end-to-end so it
MUST be always generated by each node without waiting a response from
downstream.

You can avoid OpenSIPS sending the mandatory 100 Trying but it will
cause INVITE retransmissions from the core switch. It's much better
and common the following solution:

- UA1 (core switch) sends INVITE to OpenSIPS.

- OpenSIPS replies the mandatory 100 Trying.

- OpenSIPS routes the call to UA2.

- Let's imagine UA2 is down so there is no reply.

- OpenSIPS then acts on failure_route (the winning response is a local
generated 408 Timeout) and sends a response (maybe 408, 503 depending
on UA1 failover implementation) to UA1.

- UA1 receives such response (i.e 408) and reroutes the call.


-- 
Iñaki Baz Castillo
<ibc at aliax.net>