On Tue, Apr 6, 2010 at 1:37 PM, JOLY, ROBERT (ROBERT) <[email protected]> wrote:
>> On Tue, Apr 6, 2010 at 12:28 PM, JOLY, ROBERT (ROBERT)
>> <[email protected]> wrote:
>> >> Attached is a sip trace (from QA) for a call hold ( with MOH) and
>> >> resume operation which results in no voice path after
>> resume. Problem
>> >> reported is that after the call resume there is no media path.
>> >> Looking at the pcap, the ITSP sends G729 whereas sipx
>> sends the ITSP
>> >> G711 ( asymmetric meida ).
>> >>
>> >> Note that on the re-INVITE (with SDP), sipX sends the ITSP all
>> >> supported codecs (including G729 - I need to filter this
>> out but that
>> >> is a different problem ). The phone selects G711 and ITSP sends in
>> >> G711 in the period of time following Frame 46. This is legal ( the
>> >> ITSP answered with G711 and G729)  However, it appears that the
>> >> asymmetry of the media stream causes the observed problem of no
>> >> speech path. So there appears to be an ITSP issue at hand.
>> >>
>> >> The asymmetric codec problem can be resolved again by sipxbridge
>> >> restricting all re-INVITEs responses to  to that codec
>> negotiated for
>> >> the FIRST call setup. This seems quite restrictive however and
>> >> clearly the RFC allows for multiple codecs in the response to that
>> >> re-INVITE but there is no speech path. So I am wondering
>> if I should
>> >> implement a single codec restriction on the re-INVITE response to
>> >> avoid the situation.
>> >
>> > One possible approach that would less restrictive would be
>> for the sipXbridge to implement the codec lockdown procedure
>> described in RFC3264 section 10.2 when it sees an SDP
>> containing for than 1 codec.
>>
>> Thanks but I am not sure I follow. I would have thought that
>> the codec lockdown is upto the endpoints ( in this case the
>> ITSP and the phone ). Why would sipxbridge need to take an
>> active part other than to relay the signaling?
>
> If I'm following you, your proposal already called for sipXbridge taking an 
> active part in the signaling by restricting the codecs in the re-INVITE 
> responses.  What I'm suggesting here is another strategy that would be less 
> restrictive but indeed would require the sipXbridge to take an even more 
> active role in the re-INVITE call flow.  I totally understand it if you judge 
> this approach overkill or too complicated.  I threw in that idea just in 
> case...
>
Yes you are right. I have decided to not implement that either.

In the 4.2 time frame, I will just file a bug report with
bandwidth.com (it appears to be a bug indeed) and perhaps can try for
something more ambitious in 5.0. I believe this will not be a problem
if I filter out G729 in the sdp re-offer in any case ( I should be
doing that in any case because of MOH restrictions).

Thanks for looking and commenting.

Regards,

Ranga.

-- 
M. Ranganathan
_______________________________________________
sipx-dev mailing list [email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev
sipXecs IP PBX -- http://www.sipfoundry.org/

Reply via email to