[OpenSIPS-Users] Why the best response is 408 instead of 486 when parallel forking?

[Bogdan-Andrei Iancu]

Hi all,

I'm trying to summarize the conclusions on this thread.

1) 6xx replies - I think is the second long discussion on this topic, 
but the idea is we have the possibility in TM to override it and 
interpret it in our own way.

2) priorities in reply selection (parallel forking) - based on the list 
started by Iñaki:

- 6XX (except if disable_6xx_block == 1)
- 3XX
- 401 / 407  (authentication)
- 415 (Unsupported Media Type)
- 420 (Bad Extension)
- 484 (Address Incomplete)
- 4XX
- 408
- 5XX

486,480 - should not have priority (according to RFC algorithm) as they 
do not provide additional information that may trigger resending the 
request.

Regards,
Bogdan

Iñaki Baz Castillo wrote:
> Hi, I've a UA1 that is registered in two locations, UA1-a and UA1-b.
> UA1-b is crashed and doesn't listen to SIP anymore.
>
> Other UA calls UA1 and OpenSIPS does parallel forking, so it gets 180 from 
> UA1-a.
>
> UA1-a replies 486, but OpenSIPS decides to reply "408 Timeout" to the caller.
>
> Why is 408 chosen over 486? IMHO it doesn't make sense since this 408 is 
> created by OpenSIPS itself, it's not a reply from downstream and basically 
> means "error contacting a branch".
> Since there is an *alive* branch (UA1-a), why inform the caller about an error 
> instead of replying the 486 choosen by UA1-a?
>
> This issue is common in some cases when registering behind NAT in routers that 
> assign a different mapped port each time the phone switches on (so the 
> location table has many contacts for that AoR, but most of them are dead if 
> the phone didn't de-register them when switching off).
>
>