[asterisk-bugs] [JIRA] (ASTERISK-21981) Pass-through support for Opus and VP8 formats

[Lorenzo Miniero (JIRA)]

     [ https://issues.asterisk.org/jira/browse/ASTERISK-21981?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lorenzo Miniero updated ASTERISK-21981:
---------------------------------------

    Attachment: asterisk_opus+vp8_passthrough.patch

Ok, I'm finally attaching the new patch: asterisk_opus+vp8_passthrough.patch

This new patch only provides passthrough support for Opus and VP8. The transcoding module for Opus and the placeholder format module for VP8 have both been removed, as they're unneeded.


A couple of notes about the patch:

     * The Opus SDP parameters negotiation (draft-ietf-payload-rtp-opus-00) in chan_sip.c is currently only a placeholder. Asterisk ignores what the caller provides, and when originating a call it currently sets a few parameters that may be unneeded as Asterisk isn't going to actually terminate the Opus stream. Considering the passthrough Opus support, what may be useful to improve the patch would be to store the parameters from a caller in the SIP structure, and copy them in the SDP to provide to the callee: in fact, Opus has some end-to-end functionality (e.g., FEC) that may be useful when supported by both parties. I'm not sure what the best behaviour would be when Asterisk is actually going to originate a call: I guess the correct thing to do would be to set no SDP parameter at all (caller and callee may fine tune this later in a re-INVITE).

     * Following the suggestion by Matt on the mailing list, I kept the computation of the number of samples in an Opus RTP packet in frame.c. It may be worth pointing out, though, that that method is, as explicitly documented in the code, copied from opus_decoder.c: libopus has a BSD license, is it compliant with the Asterisk license or would a different approach to the computation be needed to have it an acceptable inclusion? 

     * Considering the VP8 patch also includes a small modification to the RTP engine to support the transmission of a RTCP FIR message (to request a key frame), I also modified chan_sip.c in order to have it add a new SDP attribute (a=rtcp-fb:* ccm fir) whenever VP8 is involved. This is especially needed in WebRTC, as for instance Chrome ignores RTCP FIR messages if support for it hasn't been previously negotiated. I'm not sure this RTCP FIR message is, in the current state of the patch, also transparently forwarded by Asterisk when received from a leg: what is the current behaviour of Asterisk with respect to this when acting in passthrough mode? 

     * About the RTCP FIR message, as I anticipated this is used to request the transmission of a key frame. As such, I modified the behaviour of AST_CONTROL_VIDUPDATE in chan_sip.c to have it transmit a RTCP FIR message rather than a SIP INFO in the VP8 case. Olle expressed some concerns about this approach on the mailing list (http://lists.digium.com/pipermail/asterisk-dev/2013-May/060386.html). As I said in a reply, I wrote it that way to actually minimize the impact of my addition on the default behaviour, but of course this may need to be addressed properly.


I've tested the patch briefly with both PhonerLite (Opus) and Linphone (VP8) by placing a simple call to an Echo test application and it seems to work fine, but more testing would of course be suggested.

                
> Pass-through support for Opus and VP8 formats
> ---------------------------------------------
>
>                 Key: ASTERISK-21981
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-21981
>             Project: Asterisk
>          Issue Type: New Feature
>      Security Level: None
>          Components: Formats/General
>            Reporter: Tzafrir Cohen
>         Attachments: asterisk_opus+vp8_passthrough.patch
>
>
> Following discussions on the asterisk-dev mailing list, I create a bug for Lorenzo Miniero to merge the parts of his WebRTC codecs patch that can be merged.
> See:
> http://lists.digium.com/pipermail/asterisk-dev/2013-May/060388.html
> http://lists.digium.com/pipermail/asterisk-dev/2013-May/060419.html
> (If this is a duplicate: sorry for the noise. I failed to find it)
> Including this patch should allow for a review in time before the Asterisk 12 deadline. I marked this as "major" as I think this is an important feature missing in Asterisk right now. Supporting the formats is the least we can do.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.asterisk.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira