[asterisk-dev] Re: No complete handling of SIP Via: header?

[Samuel Tardieu]

>>>>> "Johansson" == Johansson Olle E <olle at voop.com> writes:

Johansson> The syntax here allows for multiple via parameters, but
Johansson> I've never met a UA that does that.

Well, I did today for the first time :)

Johansson> The RFC seems pretty unclear, since it in same cases talk
Johansson> about "inserting Via header values" and in some cases talk
Johansson> about "the topmost via header" not the "topmost/leftmost
Johansson> via header value". So to be safe, I would not implement
Johansson> this in a UA at all.

It has been implemented already, in a Cirpack equipment (used by Free
Telecom, the second largest ADSL provider in France and probably the
biggest SIP company in France as well). It looks like they started
using internal relays this morning with multivalues Via headers.

Here is a partial trace sent by someone who spotted the problem (on
the Cirpack side):

  - the request (from 212.27.52.5 to the final user) contains:
    Via: SIP/2.0/UDP 212.27.52.5:5060;branch=z9hG4bK-R0-0001f176-7de981fd,SIP/2.0/UDP 172.17.20.241:5063;emission,SIP/2.0/UDP 172.17.20.241:5062;branch=z9hG4bK-38DD-1FFF

  - the answer from Asterisk (from the final user to 212.27.52.5) contains
    (note the misplaced ";received=..." part):
    Via: SIP/2.0/UDP 212.27.52.5:5060;branch=z9hG4bK-R0-0001f176-7de981fd,SIP/2.0/UDP 172.17.20.241:5063;emission,SIP/2.0/UDP 172.17.20.241:5062;branch=z9hG4bK-38DD-1FFF;received=212.27.52.5

Johansson> Make sure you add SIP debug to your bug report and that
Johansson> you're ready to help us test patches.

I just need people to confirm that it's indeed a bug, not to write
patches (that I can do myself :-).

  Sam
-- 
Samuel Tardieu -- sam at rfc1149.net -- http://www.rfc1149.net/