==================================================== Multiple Interfaces (mif) WG IETF 76, Hiroshima, Japan Monday 1520-1720, Acacia East ============================= AGENDA: 1. Introduction & Agenda (Chairs, 5 min) 2. MIF Problem Statement (Marc Blanchet, 20 min) http://www.ietf.org/id/draft-ietf-mif-problem-statement-01.txt Marc> Short Presentation. Are we done? Margaret> Any comments? Hui> Most phones today have CMs inside the host and an API that can configure the behavior. This has not been clarified in the problem statement? Marc> Is this a problem or the solution? Hui> I think your argument is valid. Principally this is simply an Interface. If we had different implementations of such Interface would it be a problem for interoperability? Margaret> One thing that is that the problem statement is very specific to the stack. It does not talk about Apps. There is no standard mechanism for changing its default behavior if the app knows what it wants. Some implementations have APIs for this. Marc> In section 5 it is not far from what you want Margaret> It could be elaborated upon. There is just one sentence now. I will send some mail in the list Hui> Need more discussion about standardizing such an API Brian> There is no mention of the draft in the problem of referrals when a host wants to say to a host about a 3rd host, which is the subject of a BOF this week. When you multiply this by multiple Interfaces and multiple addresses it is even worse. Marc> Yes, we need to add more. Margaret> So it has come along a long way but we have some more work to do. 3. MIF Current Practices (Margaret Wasserman, 15 min) http://www.ietf.org/id/draft-ietf-mif-current-practices-00.txt Margaret> Short Presentation. Looking for some more implementations Marc> I do not think we will get information on the iPhone Margaret> We do not need every implementation and it does not have to be secret, it can just be based on how it works. Margaret> Presentation continues... Raj> Typically the config comes from the operators via provisioning rather than from the user Hui> But the factory places default policy. But for specific APNs it comes from the operator George> some of this is common because it is part of the 3GPP system, Margaret> but still there might need to be some defaults Margaret> Presentation continues... Hui>We would like the contributors to add more text to the document Yuri> you said that the next step will be an analysis document. From what perspective? Margaret> The next presentation will talk about the scope of that document. It¡¯s like a GAP analysis of the current problems/solutions Yuri> So then the user will come up with some approach that will make use of multi-IF more easy? Margaret> For now we are only analyzing the problem, we then need to discuss if/where any solutions need to be done. Arifumi> does the draft capture the issue of pseudo-IF and real interfaces? Margaret> In this draft we do not consider these different so if a device has physical and virtual interfaces it is consider that it has multiple interfaces 4. Start Work on MIF Analysis Draft (Chairs, 30 min) - Structured Discussion (no documents) Margaret> Presentation... Rajeev> Today if you do ifconf, you will get per IF information. and if you do netstat you will get GW info etc Margaret> it differs per implementation on exactly what you have per IF. For example some systems do not have per IF DNS servers and GWs. Rajeeve> but if you pick the wrong default router then the system is broken. Margaret> ho do I print to my local printer if I am connected to my home net and the VPN? In some systems without DNS lists per IF, you cannot get to that printer. In fact I need to communicate over the IF that I got the DNS resolution from, but no state is preserved in most systems Arifumi> what is the problem with per prefix address selection? Margaret> There is no mechanism to set up address selection per IF. The other thing is that if the app gives me a source address to use, then I should use the policy on the interface to which the address belong. Not sure if the policy needs to be per interface or per prefix Arifumi> in 6MAN we are looking to combine the policies in a single policy Margaret> Its OK to have one table but you should not restrict that all info comes from a single admin domain. Arifumi> we will be talking about that in 6MAN please come to the meeting Hui> Are you looking specifically at multihoming? Arifumi>The 6MAN work does not specifically look at the multiple Ifs. Hui> we are also doing destination selection Yuri> In google android both WLANa and 3G interfaces it has per IF configuration but if you turn on WLAN the 3G IF disappears. Rajeev> You can build that in every device Yuri> I was talking in the context of the practices document. George> so I guess this document will start describing what ¡°best practices¡± might be as opposed to ¡°current practices¡± Margaret> yes, a lot of the things we are talking about are obvious but are just not stated anywhere. Yuri> Would be nice to have a list of what things are nice to have Jari> Are these the goals of the document? Margaret>The doc would say if there is enough understanding to design solutions Jari>Overall for the project the proposed scope of the document makes sense but not only for this document. The Gap analysis should certainly be in the document. Hui> but if the problem has not been analyzed how can other WGs do it Jari> the analysis should indicate that some of the issues are already dealt with and point to specific gaps where no solution seems to exist Raj>Usually if we standardize something we tent to pick one way, but are we doing that here? Margaret> Depends on whether we think that makes sense for the specific problem Jari> yes there is no single answer to this 5. Informative Presentations on MIF-Related Work (if time allows) - Socket API Extension for MIF (Bo Zhou, 15 min) http://www.ietf.org/id/draft-liu-mif-api-extension-01.txt Bo> Short Presentation... Hidetoshi> WRT failover, the IFs have same or different IP addresses? Because if there are different addresses, then the application breaks Bo> not sure Sri> How do you define policy? Bo> This is just trying to introduce new API with MIF capabilities. The draft does not cover these details yet Hui>If there is a policy data structure somewhere else then this Api can be used to Sri> Trying to see how the socket state is dynamically adjusted Hui> The current API does not have specific elements related to the policy Sri> Without the data structure and everything else the API is not so helpful Hui> Yes other considerations should be included Teco> Maybe we should add Cisco o the current practices since they support MIF for years Margaret> If someone writes it up I will include it. Someone> the API is helpful to implementers to implement MIF. As part of the design think of feedback to applications - Inter-Technology Handoff for PMIPv6 (Hidetoshi Yokota, 10 min) http://www.ietf.org/id/draft-yokota-netlmm-pmipv6-mn-itho-support-02.txt Hidetoshi> Short Presentation ...Requests to put it in current practices document Margaret> If it is already in a product give me the information and I will put it in. Hidetoshi> Presentation continues... Margaret> wonder if we need to look at current IETF technologies to see if anything can be used for MIF T-M> VIF is well know technique but inter-tech handover does not exist. Hui>Send text to Margaret for the current practices Margaret> No, only existing products go there. It could be listed in analysis doc as part of what IETF technologies we have today. Sri> The technology is available today Margaret> in the analysis draft maybe - Multi-Homing Extensions for PMIPv6 (Carlos Jes?os Bernardos Cano, 10 min) http://tools.ietf.org/id/draft-bernardos-mif-pmip-01.txt Carlos> Short Presentation ... Hui > This presentation is based on weak model, and previous one is based on virtual interface of the connection managers. Carlos> yes, different way to do it. - MP-TCP (Mark Handley) Mark> Short presentation. Just flagging the fact that we might need Pasi> Isn¡¯t this the strong model? Mark> It is similar¡­we want to look at exactly what it is. Pasi> it is already mandated for environments where you are behind ingress filtering. 6. Wrap-up & Next Steps (Chairs, 5 min)