Re: Manage RTP and RTCP transports

The candidates between RTP and RTCP are completely separate if the two are
not muxed.


On Mon, Jun 2, 2014 at 3:53 PM, Shijun Sun <shijuns@microsoft.com> wrote:

>  Re the ICE candidates, will the list provided to the RTP transport be
> applicable to its RTCP counterpart? I agree we need to define an attribute
> to track its state if we go down this path.
>
> -Shijun
>
> *From:* Peter Thatcher [mailto:pthatcher@google.com <pthatcher@google.com>]
>
> *Sent:* Monday, June 2, 2014 3:38 PM
> *To:* Shijun Sun
> *Cc:* public-ortc@w3.org
> *Subject:* Re: Manage RTP and RTCP transports
>
> I don't think we can return void
> from  RTCIceTransport.createAssociatedTransport()  because then the JS
> can't provide it candidates, track its state, etc.
>
> Whether the rtcp Transport is explicitly tracking RtpSender and
> RtpReceiver depends on which option we choose.  In the last email about
> this topic I sent to Bernard, I showed 3 or 4 options.  That is one of
> them.  But some of the options do not require that.
>
> On Sat, May 31, 2014 at 7:23 AM, Shijun Sun <*shijuns@microsoft.com*
> <shijuns@microsoft.com>> wrote:
> Looking through recent discussions on handling RTP and RTCP when they are
> on separate transports, I like the approach in the
> RTCIceTransportController, where the UA is expected to keep track of the
> pairing of RTP and RTCP transports internally, and adding an RCTP transport
> explicitly to the controller will throw exception.
>
> A couple thoughts based on that.
>
> Do we need the rtcpTransport on the RTCRtpSender and the RTCRtpReceiver
> explicitly?  If we can track the pairing of RTP and RTCP internally, the
> interface can be a bit cleaner.
>
> To go one step further, is it possible to keep the RTCP transports as
> internal to UA?  For example, I wonder if it makes sense to have
> RTCIceTransport.createAssociatedTransport() just return void and keep the
> new RTCP RTCIceTransport internal and indicate that with a simple readonly
> attribute (e.g. boolean associatedTransportCreated, or something like
> that).  We can add a new attribute for the RTCP transport state or redefine
> the existing "state" attribute as a combined state.  A similar approach can
> be applied to the RTCDtlsTransport if this is along the right direction.
>
> Thanks, Shijun
>
>
>