Multimob (Multicast Mobility) IETF-78 Chairs: Stig Venaas and Behcet Sarikaya Note-taker: Matthias Waehlisch (including Jabber log snippets by Gorry Fairhurst) Behcet: Presented the agenda. MLD Tuning. Qin Wu: Presented draft-wu-multimob-igmp-mld-tuning-02. Hitoshia Asaeda: Regarding slide 9, does it require to install a box to do explicit tracking for IGMP/MLD? Qin: This is a role of the proxy. The current standard does not say how to do this. Hitoshi: Is this a required component to tune the IGMP/MLD protocol? Qin: It is an optional feature. Hitoshi: Good idea to mix unicast and multicast general queries by different timers, but it is impossible to do this without changing the protocol. Gorry Fairhurst: I didn't see in the draft what to do when you loose the initial Join messages from a client. To me, it was not clear whether general queries were always sent? Qin: Yes, we can use a General Query, which are always sent. Gorry: Agree with that, I didn't see that when I read the draft. Hitoshi: Presented draft-asaeda-multimob-igmp-mld-optimization-03. Qin: Are you suggesting that in case of explicit tracking the router should be aware of application type. Hitoshi: No, it is not application dependent. Qin: This feature is appropriate for wireless mobile environment. Hitoshi: This kind of networks is not discussed in this particular draft. Gorry: Is the draft talking about explicit tracking on router or proxies? And is there a difference? Hitoshi: Proxy includes router, the draft is about both. Both are described in the same way. Thomas Schmidt: What is the meaning of the axes? Hitoshi: The X-axes is a 2 hours multicast channel, duration is not important. Y-axes is number of queries. Actually only one packet is sent every 125 secs. Thomas Schmidt: Regarding slide 8, what is the binning size? Hitoshi: There is no binning, but maybe I do not understand the question. Juan Carlos: Can you explain the time scale more precisely? What is the criteria for setting these values? Thomas: The Y-axes shows messages. Does a value of 28 indicate all messages in one second. Hitoshi: This is a minute. One line is one second. It is clearer if you zoom in. Matthias Waehlisch: Why are your simulations appropriate, can you comment on the scenario? Hitoshi: This is only an example. Matthias: You should discover the solution space including upper and lower bounds.. Gorry: Why should the Robustness Variable not exceed 2? "Should Not" seems a quite strong recommendation. Hitoshi: It seems to add congestion at layer 2 with no benefit. But if this is thought to be too strong we could say recommended in cases where this could aggregate congestion. Stig: Are we ready to take one of the MLD drafts as WG draft? Thomas: Both drafts do not explain the problem clearly. It is unclear what they want to achieve. Qin: There are requirements in the charter - one sections summarises this. A number of solutions are proposed. There may be no single solution, and you may need to combine different sets of functions to satisfy the needs. Hitoshi: Our draft is clearly explaining the draft. In phase 2, we more clearly define the values for tuning. Alexandru Petrescu: With respect to the user scenarios: Do you also consider mobile hosts and mobile routers, or only mobile hosts. Hitoshi: Only mobile hosts. Alexandru: If so the time values for the frequency of the initial report may not be short enough, because the host sends MLD unsol, join etc - later messages can not take place. Hitoshi: I will think about this. Rajeev Koodli: What is the intended status of the draft(s)? Stig: Informational or BCP. Rajeev: Is the purpose how to parametrize IGMP/MLD? Is this an optimization or recommendation problem? Stig: Recommendation of parameter tuning. Rajeev: It is very hard to generalize this for all type of wireless links. In this case, it should be clear - you need data to make recommendations that are useful. Thomas: Agree with Rajeev, this picks up the discussion of Hiroshima. We have very different types of wireless links that should be considered. Rajeev: In other words, what bad things happen if you don't do the consensus calls today? - My suggestion is to wait for more data Behcet: I am not sure if we can get values for all link types. Maybe it is a bit hard to achieve your suggestion. Stig: In some cases, you can say that if you have "these" characteristics, do this. Rajeev: This seems fine. However, you should be clear on what you propose and what you are doing. Alexandru: MLD is an IP packet. We can derive values from usual IP measurements. Stig: We conclude both authors should work on their draft and provide more data. Hitoshi: If we have generalized timers that would be nice, but it is hard work. Gorry: We probably not need a perfect model, but need some recommendations with respect to the defined problems. What I'd really like to see is that the document clearly says there is a specific consideration, not that people should use 135 instead of 120 for a specific parameter. I'd prefer to see the "problem" and the "suggested remedy" and let the operator decide if they have the problem. Stig: How many people think that one of the documents is in sufficient state to be adopted? How many people think that we should wait. Stig: Sounds that we should wait. We should encourage an updated contribution from both authors and then to see whether we can then adopt these. Rechartering Discussion. Dirk von Hugo: Presented draft-von-hugo-multimob-future-work-02. Thomas: Just a comment on the categories: (a) Provide multicast routing natively. (b) Add multicast context transfer to the mobility protocols. Behcet: Presented problem descriptions by working group members. Hitoshi: Presented shortly IGMP/MLD extension proposal. Stig: Is this topic useful? How many people are interested to work on this? Alexandru: Interested in fast attachment. Stig: We only want to get an idea, which topics may be of interest. Today, there is no decision on drafts or topics. The drafts just illustrate prior interest/work and details will be up for discussion later. 8 people raised hands to be interested on the IGMP/MLD extension + 1 mic. Stig: How many are interested in IGMPv3/MLDv2 reliability wireless extension. No interest. Thomas: Presented direct routing option topic. Juan Carlos: Presented dedicated multicast LMA option. Behcet: Direct routing option and dedicated multicast LMA option are treated as one topic. Stig: How many people think that we should work on one of these topics. 8 people raised hands + 2 via jabber. Thomas: Presented multicast context transfer topic. Hitoshi: Presented PMIP protocol extension topic, which allows for seamless multicast handover. Hui Deng: Presented fast handover topic. Carlos: Presented topic on PMIP protocol extension to accelerate MAG knowledge about MN multicast subscriptions. Behcet: The topics can be grouped under handover optimizations including context transfer. How many people think that we should work on it? 13 people raised their hands + 2 via jabber. Thomas: Presented the multicast source support topic. Zhang: Presented a multicast source proposal. Stig: How many people think that we should work on multicast source support in PMIP? 14 people raised their hands. Jari: Complete the work in the current charter, half part is done. MLD/IGMP stuff is still open. Regarding charter proposals, we should also look on other people opinions. We should explore whether there is a customer who needs the work - that is something we did not hear today, but the list could provide time to allow people to respond from companies who need this. Additionally, you should cut down the topics for re-chartering to 1-2. Complete something and then move forward. We are fairly ready to make decisions in weeks rather than months. Chairs will try to get a new charter as soon as they can Behcet and Stig concluded the session.