SAVI working group meeting at IETF 77 Tuesday, March 23, 17:40 - 19:40 Chairs: Christian Vogt, Jean-Michel Combes 17:40 Introduction and Administrativa - Christian Vogt, Jean-Michel Combes Christian: BBF liaison on IPv6 security. BBF wants to avoid dad traffic between customer premises. Multiple customer premises maybe within the same broadcast domain. Alan: BBF is not asking to solve this. BBF wants to solve duplicates N:1VLAN case. DAD traffic is addressed in BBF using split horizon. Christian: what is the problem BBF that asks Alan: N:1 VLAN scenario duplicates - how to solve that. Christian: solution with proxy DAD. - will be specified in 6man. Plan is to sketch a solution and see if BBF approves Alan: need to clearly state the problem to be solved Christian: would like to move forward and respond to BBF. Status update on Framework doc: Christian: not much progress since last meeting. 17:50 SAVI for DHCP - Jun Bi (draft-ietf-savi-dhcp-01) Eric: if hub attached in the network behind a savi switch crashes, the savi device is not going to do anything. Do we have to deal with hub state recovery Jun: this draft deals with savi device recovery. Marcelo: first, mechanisms for restoring state is good. Second, we have 3 flavors of savi - slaac, send, dhcp - it would be good to have the same terminology. Questions like how to handle lifetimes, etc. - we may need to unify these. Provide rationale for this in the framework document. Would this make sense Christian: good comment. We should do this for terminology. Regarding other fine grained things - we should discuss this on a case by case basis. Marcelo: Interaction between these aspects need to be considered Christian: prioritization? Or is it something else Marcelo: I don't know Christian: wait for eric's presentation Eric: go back to binding removal: why is the host removed - local link movement Jun: if the lifetime is not over, keep the binding - good suggestion Alan: slide 22. data trigger procedure. Why is RS not used. Jun: SAVI device is going to send a DAD to the host on the link Joel: how to sort out knowable and unknowable conditions - Christian: question is what is configurable and what is not Joel: asking the user to recover a connection is not a good idea. Jun: use a trigger to indicate to the host. Eric: specific conditions for each kind of network / deployment model makes it complex. Christian: explain on ml why this procedure / pros/cons - manual configs are worth the cost Marcelo: do not agree Christian: some issues with auto config will summarized on ml Woj: clarification how does the savi device know that the host is using dhcp Jun: it is configured in the savi device. 18:10 Analysis of Open Issues in SAVI - Marcelo Bagnulo (draft-bagnulo-savi-analysis-02) Jun: discussion on ml that rate limitation cannot be handled Eric: need to software process at rate limitation and tend to lose more packets. Marcelo: dhcp packets will not be lost Jun: packet loss is not a problem for some cases Eric: is the rate limit global. Jean-Michel: optimistic IPv6 address - are they using DAD packets to start Suresh: DAD in parallel with control packets Jun: topology change can be handle by NS downstream Joel: should we depend on an IEEE protocol to handle this? Frank: clarification - nvram to store state? Joel: he is proposing not to do this Jun: for data triggered, any difference in handling between low-end, high-end switch? Marcelo: It should be a matter of configuration Jun: MUST requirement for nvram - vendors will not implement it Marcelo: qualified should - explain what will break if procedure in rfc is not followed Christian: use qualified should Eric L: we solved this a few meetings back. Triggering a binding is a strong issue Eric: lot of legacy switch is operation - servers, laptops etc. , non-compliant switches We need data triggered support. Christian: to Marcelo - send a summary to ml to start discussion. 18:50 Coexistence of Address Assignment Methods - Eric Levy-Abegnoli (draft-levy-abegnoli-savi-plbt-02) Eric Vyncke: send over human operator - change some priority Eric L: operator should configure one way Eric V: if we learn for e.g a binding through SLAAC, and then we rebind with slaac - will the timing of that rebind have some priority? Marcelo: --- Marcelo: any other changes needed in the protocol to implement this? Concern that some aspects may change Eric V: not done this yet. Cga is there Hub: what happens when there is roaming cases - wireless case. Eric L: no definition yet about when binding is removed 1. check if user has moved 2. --- Christian: fairly detailed analysis - how could this be simplified. Will these various mech compete? Eric L: needed to make a good decision Guang: are these priorities pre-defined Eric L: configurable, but default should be defined. Christian: next step will be to see how to add this to the framework doc 19:10 CNGI-CERNET2 SAVI Deployment Update - Jun Bi Christian: good progress. 19:30 SAVI for Delegated IPv6 Prefixes - John Kaippallimalil (draft-kaippallimalil-savi-dhcp-pd-01) Alan: why is this necessary - since in BBF 2 DSLAMs do not talk to each other, thus there is no interworking. John: we may need interworking to handle state recovery. Non-volatile memory is not a good solution Frank: The solution is to satisfy requirements in WT-177. BBF does not outline solution details, and this draft addresses the solution. Woj: Agree with solution, but why is this necessary. Christian: Post problem statement on savi to ml *** END OF THE MEETING ***