[OpenSIPS-Users] newtran() limitations

[Stanisław Pitucha]

2009/10/7 Brett Nemeroff <brett at nemeroff.com>:
> Can you please explain when it would be necessary to force creation of a
> transaction where t_relay wouldn't suffice?

It's really not that t_relay() doesn't suffice, but rather that
t_relay() is at the end of the script and if one process starts
working on the message and another one gets the retransmission of that
message, you process the message twice with possibly strange effects.
(for example you place 2 calls out to the gateway, even though the
phone requested only one).

The limitation I was interested in and that Bogdan verified later on
the channel is that any change to the text of the message that is done
*after* newtran(), will not be present in the failure route, or new
branches.

-- 
KTHXBYE,

Stanisław Pitucha, Gradwell Voip Engineer