Re: [rtcweb] JSEP and draft-ietf-mmusic-sdp-bundle-negotiation-03

JSEP expects that most applications will send the same SDP to the API as is
sent over the wire.

Ideally, we would have a single offer-answer exchange, and the SDP created
by createOffer and fed to setLocalDescription would work properly
regardless of what is on the other side. If that's not possible, then we
will have some constraint that toggles between the 1) fast/not-legacy-safe
and 2) slow/legacy-safe output of createOffer. We will also have to decide
on what the default should be; based on the feedback from the most recent
SIPit, where most of the legacy equipment did not even support ICE, I am
leaning towards #1.

On Wed, Feb 20, 2013 at 3:48 AM, Christer Holmberg <
christer.holmberg@ericsson.com> wrote:

> Hi,
>
> >> Good questions, for which I don't have any answers at this point :)
> >>
> >> But, IF we can agree on this as a way forward, one of the next steps is
> to look at the JSEP impacts.
> >
> > [BA] The problem is that my opinion will differ depending on whether the
> SDP in question is to be used in the API or over the wire.
>
> True. But, if SDP is NOT used on the wire, the BUNDLE document does not
> put any requirements on whatever other signaling protocol you may be using.
> You can use a protocol that ONLY signals different port numbers, or one
> that ONLE signals identical port numbers, or one that ONLY signals a single
> port number, on the wire, and there is nothing BUNDLE can do about it :)
>
> BUNDLE is only about users of SDP Offer/Answer - no matter whether it's an
> API or a wire protocol.
>
> >The "different port" formulation makes good sense to me if you are
> looking for backward compatibility with existing SIP/SDP implementations
> which are likely to send an error
> >in response to a "same port" formulation.   So if this is an "on the
> wire" question I give it a thumbs up.
> >
> >However, if you are asking me whether it makes sense for createOffer() to
> output SDP with different ports by default, when 99 percent of applications
> are
> >likely to not be doing SIP interop, I would say "no".  Then the next
> question is whether there needs to be a mechanism in the API for indicating
> a preference for different ports,
> >when this is desired.
>
> That's a good discussion topic for JSEP :)
>
> Regards,
>
> Christer
>
> _______________________________________________
> rtcweb mailing list
> rtcweb@ietf.org
> https://www.ietf.org/mailman/listinfo/rtcweb
>