> On Oct 31, 2016, at 2:36 PM, Peter Saint-Andre <[email protected]> wrote: > > On 10/31/16 12:28 PM, Philipp Hancke wrote: >> >> >> Am 29.10.2016 um 23:46 schrieb XMPP Extensions Editor: >>> This message constitutes notice of a Last Call for comments on >>> XEP-0371 (Jingle ICE Transport Method). >> >> I think its currently bad timing to LC this as the definition of the >> end-of-candidates notification is changing, see e.g. >> https://github.com/w3c/webrtc-pc/issues/726 >> >> depending on the decisions there we might have to add an ufrag to the >> example. I am tempted to say that it is a mere editiorial oversight that >> there is none currently ;-) > > Hi Fippo! > > Another potential issue was raised in a roundabout way by Jonathan Lennox on > the ICE WG list: > > https://www.ietf.org/mail-archive/web/ice/current/msg00370.html > > https://www.ietf.org/mail-archive/web/ice/current/msg00371.html > > https://www.ietf.org/mail-archive/web/ice/current/msg00372.html > > https://www.ietf.org/mail-archive/web/ice/current/msg00373.html > > Peter
Hi — yes, I was just about to e-mail this list directly about the issue. Specifically, so people don’t go have to go chasing links, this is the scenario I’m concerned about: 1. Romeo’s IP address changes, so he sends an ICE restart (ufrag B) 2. His IP address then changes again, so he sends another ICE restart (ufrag C). 3. Juliet then replies with a new set of her own candidates (ufrag X). In Jingle, Romeo doesn’t know whether Juliet intended to pair her ufrag-X candidates with his ufrag-B candidates or his ufrag-C candidates. In SIP and SDP offer/answer, this problem doesn’t arise, because Juliet’s reply would have explicitly been carried an answer to the ufrag B or ufrag C offer. (And indeed, Romeo probably wouldn’t have been allowed to send the ufrag-C offer at all until an answer came in to the ufrag-B offer.) But Jingle’s transport-info messages are unilateral, and don’t carry this semantic. Note that I’ve actually seen this issue arise in practice. It’s pretty easy to get it to happen in test cases (e.g., by forcing an endpoint to think its IP address changed once a second). _______________________________________________ Standards mailing list Info: https://mail.jabber.org/mailman/listinfo/standards Unsubscribe: [email protected] _______________________________________________
