Minutes - SIMPLE IETF74 - Wednesday 25 March 2009 - San Francisco, CA, USA Summary: * ACM Draft - The comedia use mostly non-controversial. There are open issues concerning backwards compatibility with MSRP relays. Discussion needs to happen on the list. * Multi-party Chat - The draft will go into WGLC shortly after IETF. * Intra-Domain Bridging - We discussed open issues concerning the taxonomy, role of publish, and policy composition. We ran out of time before completing discussion. The discussion was deferred to the list. Raw notes from Dean Willis: -------------------------------------- Notes on SIMPLE at IETF 74 recorded by Dean Willis Chaired by Ben Campbell, Hisham Khartabil Meeting called to order 10:30 AM Note well presented. Agenda accepted as presented Chairs reviewed status of WG, slides presented. No volunteers for Jabber scribe (note: Dean Willis tried but could not connect, it was working for him yesterday.) Topic: Alternative Connection Model for MSRP by Christer Holmberg Slides presented Rationale for this work reviewed. Item: COMEDIA for ACM. No issues have been raised. Chair polled room, consensus that the approach is not controversial. Will move forward with this working assumption. Issue: Backward Compatibility with SBC/ALG In Path Hadriel believes that the requirement to support the setup attribute in the far-end client may introduce a backward compatibility issue. He does not find this to be a problem, just wants it clarified. Modification of SDP URI was to support relays. No way to support current relays without putting new requirements on clients, and that is OK. Derek noted that there may be conditions in which this negotiation will fail with legacy systems; would rather establish up front that either comedia or a relay is required when that is the case. Ben wishes to avoid having this decision produce islands that cannot connect with each other, and that we should make every effort to make ACM work with current usage as in 4976. Hadriel assets that there is no known proposal for making RFC 4975 work without a change to RFC 4975. Consequently, there is no requirement to make this work. o Question by chairs: Do we need to support the architecture of having MSRP relays? No consensus, median split. Derek asked to revise the question: Should we revise RFC 4975 to make it work? Ben noted that the c-line approach in RFC 4975 is there for a reason, changing it would not be trivial. This probably would have consequences. Further discussion focused on this point. Discussion deferred until after Christer presents proposed solution (next slide). Issue: Alt 2 changes to RFC 4975 Noted that this change has little impact of the security properties, as they are all in the right hand side of the session ID. Issue: TLS impacts with SBC Noted that MSRP already requires hop by hop segmentation of TLS, but that this is visible to the users, whereas the breakage of e2e introduced here is transparent and possible non-consensual. Issue: TLS Certificate Matching This is primarily an impact on server-class devices. Proposed that servers could use a fingerprint approach as used for clients in 4975. Rough consensus to add the proposed negotiation mechanism and guidance, adding discussion of fingerprint usage and consequences. Question: Is this a proposal to solve legacy compatibility by requiring changes to the legacy end points: Answer: Yes. Noted that this is a media-plane question, and we usually negotiate media-plane with SDP. Endpoints can support both approaches and negotiate. Chair noted that there are more opinions in the room than there are on the list. Participants are asked to actually contribute on-list. Topic Multi-Party Chat MSRP by Geir Arne Sandbakken Slides presented Status reviewed. Are we ready for WGLC? Poll: Who has read this version: a few. Noted that the previous version got lots of feedback. Also noted that this approach does not conflict with XCON. General consensus in favor of WGLC noted. Topic: Intra Domain Bridging by Jonathan Rosenberg Slides presented Status reviewed. Open Issue: Partitioned, Unioned, Exclusive taxonomy Proposed to split three terms to 2x2 grid one/many vs static/dynamic Consensus to keep existing terminology noted. Poll: Who has read this or a previous version? (many) Open Issue: Role of Publish Proposal to add discussion accepted. Open Issue: Policy composition and union federation Is common-policy union the right model? Noted that this might create surprises to user-expectations. Need to be able to show what is visible at the root reflects what would be seen by crawling the tree, since users can crawl. Noted that the filter-composition policy seems to mostly correct for user expectations. No consensus noted, further discussion deferred to list.