Re: [Wish] Implementation report for draft-ietf-wish-whip-00

[Sergio Garcia Murillo <[e-mail]>]

You can use the ice transport apis to get username and password and only
fallback to sdp parsing in case browser don't support the apis


try {
//Get local ice properties
const local =
this.pc.getTransceivers()[0].sender.transport.iceTransport.getLocalParameters();
//Get them for transport
this.iceUsername = local.usernameFragment;
this.icePassword = local.password;
} catch (e) {
//Fallback for browsers not supporting ice transport
this.iceUsername = offer.sdp.match(/a=ice-ufrag:(.*)\r\n/)[1];
this.icePassword = offer.sdp.match(/a=ice-pwd:(.*)\r\n/)[1];
}


I am in favor of extending the abnf of the sdp fragment mime type [it is
supported by the rfc] to allo0w sending candidates at session level.
Mandating using RTP/,AVP at the m line seems really incorrect.

Not sure about changing the logic for not requiring username and password.
It's an edge case but the request for candidates and ice restart may be
received out of order and the username and password may be used for
discarding obsolete candidates, or worse end-of-candisates signaling.

Best regards
Sergio

El sáb., 11 sept. 2021 13:13, Juliusz Chroboczek <[e-mail]> escribió:

> >> My knowledge of ICE is very superficial, but perhaps we could make the
> >> data optional so that only the data available in the onicecandidate
> >> callback is compulsory?  For example, we could say that:
> >>
> >>   - ice-pwd is optional;
> >>
> >>   - m-line is optional when mLineIndex is 0.
>
> > I don't see the advantage of only putting ufrag and not pwd,
>
> The ufrag is available in event.candidate.usernameFragment.  Unless I'm
> mistaken, the pwd is not, it requires parsing the local description.
>
> > Same for the m-line: once you've done parsing, even in the case of
> > bundling it's easy to figure out what media the first m-line contains.
>
> The m-line is available in event.candidate.sdpMid, so no objection to
> making it compulsory, even though I don't see how it is useful since we
> mandate bundling.
>
> -- Juliusz
>