Closed stpeter closed 8 years ago
pthatcherg
commented
8 years ago At IETF 95, there was support for explicitly indicating which ufrag/pwd the candidate applies to. But the specifics are signaling-specific. So perhaps we should just update the text to say "SHOULD be explicit" in some way.
emcho
commented
8 years ago I am fine also adding to the spec a sample way of labeling them (the way Trickle for SIP does)
Emil
On Mon, Jun 27, 2016 at 10:11 AM, Peter Thatcher notifications@github.com wrote:
At IETF 95, there was support for explicitly indicating which ufrag/pwd the candidate applies to. But the specifics are signaling-specific. So perhaps we should just update the text to say "SHOULD be explicit" in some way.
https://www.ietf.org/proceedings/95/minutes/minutes-95-ice
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ice-wg/trickle/issues/3#issuecomment-228775526, or mute the thread https://github.com/notifications/unsubscribe/ABu5u_4QaOCYDpFwyhrP2z_po1B-63ALks5qP-g2gaJpZM4G_AmE .
pthatcherg
commented
8 years ago By the way, this came up in the JSEP work recently, that with ICE restarts being roll-back-able, the explicit ufrag is really needed for edge cases there. We think the right solution is using a ufrag on the JS object, not on the SDP (since we also need to include it on the end of candidates message, and since it would look weird being duplicated in so many lines in the current remote description).
pthatcherg
commented
8 years ago It looks like we already have a paragraph that indicates an implementation should have an explicit ufrag (or other correlation). Great! So, I made a PR to add some examples:
emcho
commented
8 years ago resolved
@emcho and @stpeter had an editors' discussion (see https://github.com/emcho/trickle-ice/issues/3) about how to specify the handling of ICE restarts in the context of Trickle ICE. One possible approach would be to send a new ufrag and pwd without triggering a full offer/answer exchange, because Trickle (and ICE itself) can be used with application protocols that are not tied to O/A. There is an SDP dependency on the a= line, but we've disengaged Trickle ICE from SDP, too. Also, if the sender doesn't include a media description then a new offer might not be necessary anyway. I'm hoping that @emcho can propose text...