Minutes for DISPATCH at IETF-91
minutes-91-dispatch-1
Meeting Minutes | Dispatch (dispatch) WG | |
---|---|---|
Date and time | 2014-11-11 19:00 | |
Title | Minutes for DISPATCH at IETF-91 | |
State | Active | |
Other versions | plain text | |
Last updated | 2014-12-19 |
minutes-91-dispatch-1
Minutes from Dispatch at IETF 91 1) We clarified Keith's questions aboutif we had previous dispatch one or two drafts to MMUSIC. The conclusion is that both draft-ejzak-mmusic-data-channel-sdpneg draft-ejzak-mmusic-msrp-usage-data-channel were dispatch to mmusic. 2) Deadlines for IETF-92 will be sent to the list after we get the deadlines for drafts at IETF 92. 3) Discussed Correlation of multiple responses of forked INVITES in Back to Back User Agents Conclusion: Right now it was not clear enough for people to clearly decide the scope and value. We have asked the authors to continue on the individual draft on the dispatch list. Raw notes below -------------------------------------------------- Below are the notes courtesy of Brian Rosen Agenda bash (Chairs) Dates will come on the list Keith Drage: Doubt on SDP for MSRP draft dispatch Chairs: Dispatch is good with your suggestion, up to ADs. Both docs were dispatched Correlation of multiple responses of forked INVITES in Back to Back User Agents (Roland Jesske) Keith: No-fork is not a directive, only advisory <scenarios> Jon P: confused on how forking proxy returns candidates to B2BUA Keith: Need a better reason to have document than implementations are misbehaving Keith: Second scenario not reasonable, No-fork directive is only advisory Paul K: Not sure if these are multiple responses in single early dialog, or more than one Chairs: So do we need a real use case other than broken hardware Jon P: Yes, need something more than broken implementations Paul K: The B2BUA will itself be difficult to get right Christer: Guidelines on how something should be done helps interoperable implementations. Early media with multiple dialogs has always been a problem. Adam R: Use cases show all we can do is move the problem. Clarifications might be okay, no new protocol mechanisms. Robert S: Not just user agents misbehaving, also what the network does to get things to that user agent Keith: Is there a problem with existing specifications (imprecise specification) that we could fix? Chairs: He wants more deterministic behavior Chairs: Should we improve the document to prevent bad behavior or describe new behavior to fix bad implementations? Paul K: Not seeing any new protocol mechanism, it's best practice or advice Robert: Or document that if you do it wrong you won't be able to make it work Christer: If this goes to STRAW, it's not new mechanism, just explaining how to do it right Robert: Agree that a good survey and analysis when there is logic correlating multiple dialogs to one dialog. We have a notion of a "conference servers", with protocols. Could use a lot of these ideas. Jonathan: Not clear if this is lazy implementation or engineering constraint. If it's the latter, need some analysis and possible alternatives. Keith: Not happy describing a box that turns multiple dialog into one. Could see advice on handling multiple dialogs Christer: We have 3261 (From-tag),... Chairs: Proposal - write an individual draft so we see how this is proposed to work, then decide what to do and then we can make a dispatch decision. Keith: Anyone can do that, so no problem. Describing a box isn't good IETF work (in a work group) Robert: Rewrite draft as a problem statement and come back? Chairs: Problem statement and proposal for how to solve it Robert: <To Roland> Do you understand what to do? Roland: I thought I did that. How deep should I go? Robert: I agree, people available to help, What list? Chairs: Dispatch Paul K: Responding to Keith, we do this kind of stuff in STRAW Chairs: STRAW does more than explanations Paul K: But the point is we do that kind of stuff Keith: Not proven to me that this within the scope of IETF Chairs: Agree, that's why we want another draft, then decide Paul: Existing doc has much of what you want Chairs: Yes, but we want more Richard Barnes: "It's not clear" has been said a lot. Better problem statement and sketch out straw-man solution ------------------------------------------ Some extra notes from Cullen bellow Jon Peterson pointed out use case just did not look like sip. Keith Drage pointed out we need uses cases for non broken equipment. Keith pointed out presentation seemed someone confused about what no-fork directive did. Paul K and others questions why this is better than fixing it Christer - if we took on this, we would need to study what information we want to provide and what is ok to loose Adam Roach - the best we can do is move the problem around in the network and that does not solve anything. What we need is a clarification document of what can happen. Robert Sparks - it's not just UA behavior, but also to what the network is doing to get things to the UA Keith D - is there a problem to solve here. Is there some existing UA RFC that is not clear enough. Would we get more deterministic behavior to say Keith does not see new protocol coming out of this Christer something here would help people design network Sparks - it would be useful to have a survey and analysis of what existing RFC have to support this type of arch where one correlates multiple dialogs to a single dialog. Keith - we have forking in SIP. Hesitant to add a spec that says how to write a device that converts multiple dialogs to one. Would be willing to say if you have a box that has multiple dialogs, here is how you correlate them together. Should not say how to convert it back to a single doc.