[Devel] append_branch - in failure route

[Klaus Darilion]

On Wed, February 7, 2007 11:10, Bogdan-Andrei Iancu said:
> Klaus Darilion wrote:
>> Hi!
>>
>> As we have clarified the normal append_branch behaviour maybe we can
>> clarify the meaning in failure_route too:
>>
>> failure_route[1] {
>>   # send to voicebox
>>   rewriteuri("sip:vm at server.com");
>>   append_branch();
>>   t_relay();
>> }
>>
>> I suspect that the new URI will be written into RURI (branches[0]). Then
>> append_branch() copies the URI from branches[0] into branches[1]. Then
>> t_relay sees 2 branches: branches[0] and branches[1]. But as branches[0]
>> is already "dead", only branches[1] will be releayed.
>>
>> Is this correct?
>>
> yes this is correct - there is no special handling of append_branch() in
> failure route. What is important to know is that the branch[0] in
> failure route is the winning branch of the transaction and and inherits
> all its attributes. So, if you had force_socket(sock1) for branch x
> (before relay) and this is the winning branch in failure route, the
> branch[0] will have the forced_socket "sock1" which will be again
> inherit by the new branch after append_branch().

Thus - in failure_route - I always have to
 1. change URI to new destination
 2. append_branch().

If I would do it the other way round (1. append_branch, 2. change URI)
then I would again call the same URI.

Correct???

thanks
klaus


>
> regards,
> bogdan
>