Minutes of Multimob - IETF 83 Taken by Akbar. Behcet introduces the agenda. Akbar takes the official minutes. Seil is on Jabber. Pete is on Jabber Behcet: Status of WG drafts: Good progress so far. IGMP/MLD Tuning in IESG last call. Two new WG drafts since last meeting. Tunnel Convergence I-D and Source Mobility I-D. Note Well mentioned. 1) Juan Carlos presents Multicast Routing Optimization WG I-D. - Shows changes that were made since document became adopted as WG draft. - Multiple changes. Mainly editorial including restructuring. - Main concept is that MAG can be an MLD Proxy or a Multicast Router. - Comment from Behcet: Current MAG is Proxy only in WG. Can it really be a multicast router as well? - JC: This point will be re-visited in the I-D. - Comment from Mike (Huawei): Is this probelm documented somewhere. - Comment from Thomas: It can be a multicast router. Previous discussion on WG list proposal was wrong. - Comment from Stig: It can be a multicast router. - JC: MTMA serves as mobility anchor - JC: Suggested next steps. Some more re-organization to describe by node instead of functionality. And provide more details. - Behcet: Editorial comment. I am concerned about name of routing optimization. Should be more tunnel convergence. - JC: Name came from charter milestone names. - Behcet: Also there is another new draft called route optimization - Stig: This does solve convergence issue, but discussed different models so it is not just about convergence. Thinks title does not have to be solely convergence. - JC: More work will be done by authors and requested input from WG - Stig: Should discuss more how to choose or get the content. A little more discussions on policy would be good. - JC: We have a small section and plan to put more. For example, use SIPTO like solution for policy. - Stig: Talk a little more about policy problem. - Behcet: We feel I-D is in good shape and hope that it will go to last call soon. - Thomas: There are still outstanding questions on list. Not ready for last call yet. 2) Thomas: New WG draft on Multicast Source Mobility for PMIP - Addresses 3 different scenarios. Pragmatic base line approach for listeneres. Direct Multicast, and optimized solution for mobility. - Gave document history. Merger of two previous drafts. - Described base solution. Gave caveat that proxy should check for subscriptions or could encounter flooding. - Stig: This is just basic proxy functionality. - Thomas: Yes. - Described PIM-SM direct connect problem. Solution is to configure tunnel interface with border bit that ends PIM domain. - There are several efficiency issues. - Described direct multicast routing. Has correspondance to solution that Juan Carlos just presented. - Described alternative approaches. - Stig: There might be some policies involved and would be good to explain this. - Thomas: That would be a mixed deployment case. - Described optimized source mobility. This is a work in progress. - Juan Carlos: Is MUIMP based on source or destination address? - Thomas: This is work in progress. Current thinking is separate on group address. Separting on source address is not obvious. - Summary: This is handling source mobility in the basic solution as a first step. 3) Carlos presents PMIPv6 Multicast HO optimization - Described proposal characteristics (generic, simple, robust, precise, light extensions to PMIPv6, minimal develoment and deployment impact). - Behcet: What does it only extends base PMIPv6 protocol mean? - Carlos: No extension on the MN, only PMIP extensions. - Described proactive handover case. - Thomas: Does LMA have to store all states for all time? - Carlos: Only for certain cases. Not always. - Described reactive handover case. - Thomas: This contradicts previous statements. - Carlos: No, and describes some points ... - Described changes from last version and next steps of adoption as WG document for a basic multicast handover optimization mechanism in PMIPv6. - Thomas: Has limited feelings about solution that claims fast handover but is actually slower then base case. - Carlos: Reactive case is when this happens and it is only to cover all possible scenarios. Reactive solution will not be slower then base case. - Stig: Might it be worth, doing the proactive solution, and in the reactive case count on MLD only? - Carlos: This is a possibility. We did simulations and even for reactive case we sometimes saw benefits. We may add numbers in an appendix. - Marco: Reactive case is always tricky. Signalling may not always be faster. Is inter-MAG something to specifically avoid? - Carlos: We do avoid inter-MAG signalling by going through LMA. - Sandra: Traffic may be affected by waiting for response from previous node? - Carlos: We can think of that. Good point. - Behcet: We asked Marco to make comments which have been posted on list. Since the process has been going on a long time. In this meeting we would like to make a decision if possible. Is the WG ready to adopt this I-D as a WG draft? We could go for experimental status to prove the solution. - Behcet: Asked for show of hands for adoption. (About 10 people) - Behcet: Who is against? (None). - Behcet/Stig: Good support for WG adoption. Will confirm on list. - Behcet: Possible new charter item for this group will now be discussed. 4) Dapeng on DMM - There is a new WG on DMM with a lot of new I-Ds. - Provides a classification of DMM I-Ds (dyanmic anchoring on PMIP, control/data plane separation on PMIP, CMIP based, routing based, Loc/ID separation). - Provides diagram and brief description of each of the classes of solution. - Behcet: Is reference to NAT to IPv4 or IPv6? - Dapeng: IPv6. - Stig: Is this stateless? - Dapeng: Yes. - Mohan: How is this mapping between Loc/ID done? - Dapeng: Mapping system is out of scope in the I-D - Wein: Clarification that for control/data plane separation the first 2 drafts are different from last 3 drafts. Anchor relocation is different between them. - Dapeng: Thanks for the clarification. - Thomas: Is there any new challenge for multimob WG? - Dapeng: Next presentation will cover that. 5) Dirk on DMMM (Multimob possible future work?) - So far multimob focused on PMIPv6. What about DMM? Would base solution apply to flat infrastructure? - DMM is still being defined in the new DMM WG. - New networks evolve toward flat architecture due to mobile data traffic growth. Mainly video -> efficeint data dissemination -> multicast - Thomas: Multicast is typically not a single source. - Dirk: Ok. Meant less sources then receivers (not single). - Way forward for multimob: assume generic dmm components. Multiple anchors at the edge and not central entitiy. - Described possible assumed architecture for DMM Multicast. Is optimization broken in this architecture? - Stig: DMM is just starting. How long will it take to pick solution for DMM? Can multimob work on this until DMM is more solid? - Dirk: It may be true that we have to wait for DMM to progress. - Marco: There are impacts to multicast listener. Starting with DMM analysis in this WG is valuable. - Seil: Multimob should look at how our base solution can be applied to DMM. Use case is well understood. - Thomas: It might be useful to do problem analysis of DMM that is independent of individual solutions. - Sergio: This is exactly what we did in use case analysis. - Behcet: First do scenario analysis without going into multicast solution for DMM as that is not yet defined. We will discuss with our new AD. Also should discuss over list. - Stig: Typically we worked on charter items, but we have also worked on outside charter items. We can try to keep doing this for DMM. - Behcet: Asked WG for show of hands of people who know DMM from before this meeting. (About 80% of group raised their hands). 6)Mike on IGMP and MLD optimization in wireless and mobile networks. - Wireless can be WiFi, WiMax, etc. - Stig: We have an existing document on tuning. Is this on changing the protocol? - Mike: Yes, the existing document covers timers. We are not asking for this I-D to be adopted as this is outside Multimob charter. This covers extensions to MLD. - Aim to minimal changes to MLD to work better in wireless. - Lists 6 options. - Switch between unicast and multicast general queries. Enable the router to swtich according to actual network conditions. Need explicit tracking to do this. - Described General query supplemented with unicast general query. - Described General query suppresion with no receiver. - Stig: General unicast query is more robust. - Mike: Yes. - Described Tuning response delay accorindg to link type and status. - Behcet: Point to point link can be a big issue for us. PMIP work for pt to pt. But wifi is not pt to pt. This is something to explore more. - Described Triggering reports and queries during handover. - Thomas: If you don't have a pt to pt link, does the AR know about the arrival? - Mike: Yes, that is good point. - Behcet: Does this look like a good topic for the WG to explore further? Work on IGMP/MLD behaviour optimizations? Asked for show of hands. - Charles: Are you asking for specifically for extensions? - Stig: This issue came up before. We could ask for changes to the charter. This proposal was not for new messages but for new behaviour. - Behcet: Asked again for show of hand. Who is in favor of doing this work? (7 hands go up). Anyone against? (No one against). 7) Thomas on Possible Future Work III: Charter Issues - Describes current charter wording and notes that: Fast handover, vertical handover for service optimizations, inter-provider/provider-independent approaches. - Mike: Was LISP ever discussed in multimob? - Thomas: No. - Behcet: Was LISP discussed in DMM? - Behcet: Do people agree with Thomas' proposal? - Jari: Thanks for pointing out the charter language needs fixing. The point of charter wording is to focus the group and prioritize the work. Do we need to look beyond PMIP? That is the key question. - Behcet: Key point was to support PMIP and not CMIP. - Jari: We can review the charter language. Key point is whether there is industry support for any new work. - Thomas: Main point is that there is editorial problem in the charter, but the bigger issue is do we want to go beyond PMIP? And currently the charter is contradictory on this. - Behcet: If we work on DMM we will have to go over whole charter. - Jari: Intent of charter was to work on PMIP and all extensions thereof. But not CMIP or LISP. - Behcet: Proxy FMIP would automatically be included in chater? - Jari: I think so. - Behcet: Thanks. We adjourn a bit early. Please sign blue sheets. Thank you!