[Blink] FW: [SIP2SIP - Defect #2145] (New) Test VUC and Contact Fail with SIP Error Code PJMEDIA_EINVALIDPT

[Saúl Ibarra Corretgé]

On Aug 20, 2013, at 12:34 AM, David Frankel wrote:

> ZipDX does now "tentatively" support Opus.
> 
> When Blink offers Opus in the SDP, ZipDX responds back with Opus in the
> answer.
> 
> For some reason, Blink then complains:
> 
> SIP core error: Could not parse SDP for audio session; Invalid media payload
> type (PJMEDIA_EINVALIDPT)
> 
> I do not know what Blink doesn't like about the SDP coming from ZipDX;
> somebody will have to help us figure that out. Here's the SDP:
> 
> v=0
> o=telurix 18754 3585932952 IN IP4 76.74.151.123
> s=-
> c=IN IP4 76.74.151.123
> t=0 0
> m=audio 12980 RTP/AVP 108 96
> a=rtpmap:108 opus/48000/2
> a=rtpmap:96 telephone-event/8000
> a=fmtp:96 0-15
> a=ptime:20
> --
> 

Hi David,

I fixed this in Blink, it will be part of the next release. I hadn't realized it was you who sent the report, so I'll add some details which may help interoperability :-)

Basically the problem is that while Blink offers this:

a=rtpmap:108 opus/48000

ZipDX replies with this:

a=rtpmap:108 opus/48000/2

Note the /2 at the end, that's the important bit here. That is supposed to indicate the number of channels, and as of the latest version of the opus payload draft (http://tools.ietf.org/html/draft-ietf-payload-rtp-opus-01) says, you are correct and there must always be a /2 there.

Now, this wasn't the case in the first 2 versions of the draft, so out of 5 implementations I tested, ZipDX is the only one doing it the new way. Two other client implementations fail when calling ZipDX as a result. (not counting Blink since I just fixed it).

This results in a codec mismatch (effectively it would be the same as if I offered g722 and you replied with g729) so the session is ended.

I don't know the reasoning behind this but I find it very problematic: on every other codec (such as L16) that /2 indicates stereo mode. In opus, however, since parameters are negotiated inband and with stereo=1 format parameter, there is no real need to specify it AFAIS. I guess the draft is being discussed in AVTCORE, maybe we can get some answers there.


Regards,

--
Saúl Ibarra Corretgé
AG Projects