Multimob Minutes:

 

1) Chairs Administrativia

- Bechet - We will cover WG drafts and would like to advance to WGLC if possible (authors should indicate during presentation the status).  Also some new I-Ds.

 

 

 

2) Juan Carlos (JC)- Tunnel Convergence WG draft update

- JC presents "Multicast Routing Optimization" slides

- Thomas- Should correct IPv4 support text in document

- JC- Agreed, will do so

- Thomas- Again IPv4 issue on Pg. 5/5 as descriptor def'n is for MLD but must be added for IPv4 IGMP

- Thomas- Direct routing is not well described in text.  The I-D should upgrade text.

- JC- We may benefit from other I-Ds in Multiple Proxy Upstream

- Hitoshi- Most parts of I-D are stable and in good shape.  Including PIM discussion may distract and confuse the issue.  Recommend to exclude info on PIM and base it on MLD Proxy.

- JC- We will consider this.

- Thomas: I-D started with MTM-A and Direct Routing and now is doing something else.

- JC- That is not accurate.  We are solving issues but not asked to solve PIM issues.  That is what Hitoshi is also saying.

- Hitoshi- Having Direct Routing is okay, but discussing PIM is not needed.

- Bechet- We need to make the I-D as Proposed Standards

- JC- We need some changes so we need to make it Proposed Standards

- Brian- Should make it Experimental.  Avoid hurdle from other ADs due to running code requirements.

- JC- Okay.

- Volunteers to review I-D (soon)- Hitoshi, Akbar will review the draft

 

 

3)Thomas- Source Mobility WG draft update 

- Thomas presents "Source Mobility" slides.

- Hitoshi - Doesn't cover Multiple Upstream Proxy as you claim

- Thomas - Let me present first

- Bechet- What is the relation to 6224?

- Thomas - 6224 was informational, here we extend function of proxy so there is a change.  We merged idea from Sri

- Carlos - Supporting multiple upstreams is something we want to do.  But there are other drafts.  So do we want a more general solution then you have.

- Thomas - It is a bit late to bring this up.

- Carlos - I never understood that you were handling this.

- Stig - (As an individual).  Using multiple upstream interface is not just for source mobility.  It makes sense to describe separately how it should work.

- Thomas - We have described this for several IETFs.

- Stig - There were other drafts talking about it.

- Thomas - You are right that multiple upstream proxy is beyond source mobility

- JC - In the PIM group they recommended to define MLD Extensions in PIM WG with close coord with Multimob

- Stig - Do you describe how to construct table in section 3.1?

- Thomas - We just assume that it is a given.

- Hitoshi - This kind of discussion is completely beyond source mobility

- Georgio - Kind of filter used for table needs to be described or may cause overloading.

- Thomas - We are assuming simple core cases.  Can be configured when setting up the system.

- Stig - If you have a single instance with multiple upstream interfaces, how would it work?

- Thomas - That is already standard MLD proxy operation

- Stig - Depends on how you implement it

- Stig - Ok, all local upstream traffic is shortcut

- Brian - (Speaking as MLD Proxy co-author).  You are both talking about implementation details.

- Thomas - Having association of upstream and downstream interfaces with isolations is the same as 6224.

- Brian - If you are associating upstream/downstream with Policy info then how does it work?

- Thomas - Just add a filter then.

- Hitoshi - Multiple interface support should be excluded not improved.

- Thomas - We could have this done in other drafts, but we started this work and can discuss multiple drafts

- Carlos - Can you explain what you mean about info being repeated in other drafts?

- Hitoshi - We discussed multiple interface support in Magma, but to
make the proxy spec simple, that discussion was stopped. So you did
not start it.

 

 

4) Carlos - Fast handover WG draft update

- Carlos- Presents slides

- Bechet- Is performance comparison simulations?

- Carlos- It is analyatical

- Thomas- I sent reminder on email about my previous comments on IPv4 and relation to transient binding, and does not cover unicast.

- Carlos- Transient binding for unicast is different problem.  We are just covering multicast.

- Marco- Can be either early acquistion or late binding.  So it is not appropriate to keep bringing up this issue against this draft.

- Bechet - I agree with Marco.

- Georgio - It would be good to have an overview of the protocol at the beginning.

- Brian - Do you have requirements in these documents?

- Bechet - If this is a concern then we should check and get back to you

- Brian - How will you know when the protocol is functional (for all the documents in this WG)?

- Carlos - In our case we took requirements from Charter and did analytical analysis.

- Brian - That is okay for this document

- Brian - We need to document requirements and discussions with operators in the respective documents which is very helpful for IESG reviews

 

 

5) Carlos - Extension of the MLD proxy functionality to support multiple upstream interfaces

- Carlos - Presents slides

- Thoams - What do you mean by remote and direct subs in basic source mobility?

- Carlos - We can clarify that if it is not clear

- Daipeng Liu - Are you assuming 2 interfaces simulatenously?  Is this possible on current end points?

- Carlos - This is for routers (MAGs) and not end points

- Georgio - Have you listed the requirements?

- Carlos - Yes it is in the draft

- Thomas - I have problems with the categories in the table {Summary of needed functionality per scenario)

- Carlos - If something is not clear please write to the mailing list

- Stig - What does direct subscription mean in the Source case?

- Carlos - On pg. 9 of the backup slides we cover this.

- Stig - So direct means local?

- Brian - Should describe more clearly how proxy works with any combination of sender/recevier and upstream/downstream

- Stig - If we generalize, the proxy hsas two upstream interfaces and doesn't care what the upstream node is

- Thomas - This may be overdiversified, and you often switch perspective from mobile or receiver which is confusing

- Carlos - We can improve clarity

- Georgio - Regarding upstream is it only multicast or also unicast?

- Carlos - Scenarios are PMIP but we are not limited to tunnels

- Georgio - Then you need description of these types of scenarios

- JC - Good to consider options in proxy, just want to see if we need to address in next version of our other documents?

- Stig - It depends.  Issue has to be addressed, but I can't say which document should have it. Useful also to discuss in PIM WG about multiple upstreams.

- JC - I was not suggesting putting work in other document but just to consider it.  You mentioned previous work can we get access to it.

- Bechet - This was discussed in Magma

- Hitoshi - Basic concept of multiple interfaces was discussed in NAT box but increases complexity.  We had one explicit draft that supports selection mechanism of upstream interfaces.  Multiple upstream interfaces is really trivial and not Nobel prize concept. My personal opinion is that your draft is really interesting and support continuing work but orient it towards requirements draft and not solution.

- Carlos - Thanks, our interest is to see if WG finds this work useful and don't care about which draft it is in.

- Thomas - I will be interested to see previous work in Magma and authorship.

- Brian - I want to clarify a process issue.  A WG document is consensus of WG and does not mean that the authors have copyright on the content.

- Stig - As a PIM chair, it would be good to have a discussion there about multiple upstreams and impacts to multicast.

- Brian - There are competing drafts and doesn't mean it will supersede Thomas' document but also doesn't mean that competing drafts are not allowed.

- Georgio - Maybe having design teams can be used to avoid these authorship problems?

- Stig - It will be good to continue investigation but maybe too early to adopt.

- Brian - They will present in PIM, here it is FYI for now

- Hitoshi agreed to send a pointer to the previous Magma work
 

 

6) Hitoshi - Multicast Routing Optimization by PIM-SM with PMIPv6

- Hitoshi presents slides

- Stig - You mention GRE tunnel but it could be any tunnel?

- Hitoshi - Yes, in general, but except IPv6-IPv6 tunnel as it is used
  for regular PMIPv6 unicast transmission.

- Hitoshi - Example 1 should be IPv6.  Small mistake.

- Seil - How do you distinguish between localized and direct routing (on pg. 8)?

- Stig - You are describing standard PIM operation in certain network topology.

- Hitoshi - Yes, can give hints to operator how to setup tunnel configurations.

- Georgio - How is the hash value generated?

- Hitoshi - For Cisco, we have 3 hash values

- Stig - But all this is standard implementations

- Brian - I am glad that you said this is standard PIM.  Maybe consider whether more detailed operational recommendations would be worthwhile

- Bechet - This draft has value if it solves tunnel convergence problem

- Hitoshi - JC's draft is teh base solution for tunnel convergence problem.  This draft is an optimization and works on other issues as well.

- Stig - There is some deployment guidance in this document for PIM

- Seil - This has some value for tunnel convergence problem

- Thomas - Deploying PIM in access network is one issue.  The other issue is in source.  And I don't want to repeat my old statements.

- Bechet - Then continue to work on this.  Please revise based on comments you received.

 

 

7) Seil - DMM Multicast Use Cases draft update

- Seil presents slides

- Georgio - Since DMM is still being defined maybe it will be too early to start this work.

- JC - In DMM we have a set of requirements, but we have not gotten into solution (client/network/distributed/etc.) so are you going to consider everything?

- Bechet - Do you support this work?

- JC -  I think it is early to do this here

- Seil - This is not about solution.  We need to wait for DMM unicast solution but we can still look at multicast evolution.

- Marco - I agree with your work flow starting with analysis.  But it is too early to start talking about solution as DMM unicast is not yet designed.

- Daipeng Liu - I share concern with other people.  

- Carlos - I also share previous concern but use cases and requirements are okay

- Stig - (As individual).  I think it is too early.  Better to take part in DMM WG and try to influence them there for multicast support.

- JC - Are you suggesting to add multicast requirement in DMM?

- Stig - At least bring it up there

- Brian - Yes, you should be talking to them.  DMM is still early so if people who are interested in multicast are there then it will be good.  I am also the AD there and can influence the DMM WG.  So take this draft there.

 

 

8) Chairs - Handover draft discussion

- Bechet - We have 2 additional drafts on Fast Handover

- Thomas - Why was our Fast Handover draft rejected when there was good support on the list (than previous drafts)?

- Brian - It is up to the WG chairs to decide consensus.

- Stig - We can re-look at the issue

- Brian - I will look at this and send an email to the mailing list