[OpenSIPS-Users] Hairpin routing and Loop avoidance

[Iñaki Baz Castillo]

El Jueves, 25 de Junio de 2009, Iñaki Baz Castillo escribió:
> Some Asterisk's developers involved in fixing this issue have proved not no
> have enough knowledge of SIP (specially in some core sections of RFC 3261)
> as you can realize by reading some comments in the bug report. Sometimes
> they prefer to invent "how SIP works" instead of reading the RFC (i.e.
> CANCEL To_tag in parallel forking, comparision of SIP URI's...).

And they insist on their errors even if they have been well explained by 
others.

Take as example the following extract from the comments:

- stephen_dredge:
You can potentialy get a few strange things happening if the proxy does not 
behave correctly. For instance openser will incorrectly send a intrium '200 
cancelling' response to a CANCEL with a new tag thus creating a new dialog.

- miconda:
Not sure that it was fixed by the latest patch, but the above statement is 
wrong.
An stateful proxy has to reply back 200 OK to any CANCEL that matches an 
ongoing call and then CANCEL all the branches associated with that call. Have 
in mind that the stateful proxy can do parallel forking, so it has to CANCEL 
multiple branches. It cannot send back the replies from all branches. Please 
refer to RFC3261 sections 9.2 and 16.10

- stephen_dredge:
yes you a right a stateful proxy should send a 200 ok to a cancel if there is 
a existing dialog, but it shouldn't create a new dialog/context by creating a 
new to tag. It should use a existing dialog for the response.


This is pathetic. stephen_dredge is 100% wrong in his last comment. The 200 ok 
for a CANCEL ***can*** contain a new To tag, for sure!


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