MIF Working Group Meeting (IETF 85) The MIF working group was held at IETF 85 Vancouver at Friday, Nov 9 at 9am Chairs: Margaret Wasserman and Hui Deng 1. Preliminaries (5 mins Chairs) - Blue sheets - Note takers: Stuart Cheshire and Peer Azmat Shah - Jabber scribe: Michael Richardson - Agenda 2 Document status (Chairs, 5 min) Hui: goes over the agenda and the status of working group documents. Erik Kline: Do we have implementation experience? Is there IPR on this? Hui Deng: There are two implementations, and one IPR claim. 3. draft-ietf-mif-api-extension-00 (Ted, 5 min) Sheng (huawei): Not sure how much this document can solve source address selection issues Ted Lemon: There was a good presentation in homenet about multiple routing tables Sheng (huawei): This is okay for routers, but it's unreasonable to expect hosts to have routing tables Margaret Wasserman: Every host needs some kind of routing table, because it has to have a place to put ICMP redirects 4. draft-ietf-mif-happy-eyeballs-extension (Dan, 10 min) Zhen Cao: Could we use this for DNS server selection? Dan Wing: Yes, you could do that. Ted Lemon: DNS servers should only be used in the provisioning domain that configured it Dan Wing: Agree. We may not want to look up an internal host name at an external DNS server Margaret Wasserman: With DNS you can get a valid response (e.g. NXDOMAIN) which nonetheless may not be helpful. Stuart Cheshire: This issue applies to other things apart from just DNS. A TCP connection on one interface may connect to the correct web server, while a TCP connection on another interface may connect to a captive network portal page. Both connections succeed, but don't produce the same result. Hui Deng asked why don't we propose a mechanism to use multiple interfaces. Margrett requested from the group members that whether we should consider the DNS and other applications to be included as part of this document. 5. draft-ietf-mif-dhcpv6-route-option-04 (Margaret, 30 min) Ted Lemon: This document is wasting the group's time Lorenzo Colitti: What is the process to have it removed from the charter? Margaret Wasserman: There would have to be WG consensus to ask the AD to remove it. Margaret asked for a show of hands: 11 remove from charter 14 not remove from charter No clear consensus to remove it Andrew Sullivan: The people who want it removed should make their arguments on the list Ted Lemon: On Issue 18: The offered use cases are fictitious, and not actually from any real customers that actually have those use cases or actually want DHCPv6 route option as a solution in those cases Margaret Wasserman: Asked for show of hands for removing entire use cases section and trying again 16 try again 1 keep current use cases section Issue 9: This option disregards the RFC 5505 guidance on fate sharing Konstantinos Pentikousis and Alexandru Petrescu volunteered to assist the draft authors with the rewriting of the use cases Erik Nordmark: Add paragraph explaining why this option disregards the RFC 5505 guidance on fate sharing Margaret Wasserman: Asked for show of hands 8 Add paragraph 1 Do not add paragraph Issue 5: Only one default router? Brian Carpenter: More than one "default" anything makes my head hurt Lorenzo Colitti: There can me more than one router associated with a host's default route Erik Nordmark: This term comes from RFC 4861 Dave Thaler: It's reasonable to have more than one default router Margaret Wasserman: Asked for show of hands 20 support multiple default routers 0 do not add support Issue 6: Lifetime is not 16 bits Indeed, why is lifetime there at all, since the DHCP lease already has a lifetime of its own? Margaret Wasserman: Asked for show of hands 4 remove route option lifetime and depend on lease lifetime 5 do not remove lifetime Margaret Wasserman: Asked for show of hands 7 Change to 16 bits 0 Leave at 32 bits Issue 7: Separate specific routes from default routes No objections to closing this issue with no changes Issue 8: Include default router MAC address 1 Add MAC address 15 Leave it with no MAC address Erik Kline: Do we have to update Node Requirements RFC to say that RA-less networks are to be expected? Finally, Margrett discussed the next steps as she will confirm the consensus calls on the mailing list and then to update the document as per the consensus decision. 6. draft-reddy-mif-dhcpv6-precedence-ops (Tiru, 10 min) Margaret Wasserman: It seems like a good idea to have more intelligence at the router that knows what is going on Lorenzo Colitti: This makes DHCPv6 mandatory even for hosts that are just using SLAAC Lorenzo Colitti: is this a multiple interfaces. Hui: MIF is chartered not only multiple interfaces, but also multiple provisioning domain. Lorenzo: where is multiple domain for this diagram. Hui: to the right side of access router. Some discussion how to revise the diagram to show correct topology, current there is only DHCP server in the diagram. 7. draft-seite-mif-cm (Carlos, 10 min) Dapeng Liu: Why is it important that user gets same experience on laptop and mobile phone? the interface between MIF API and application shouldn't be cut. this need to be alignment. MIF API does not need to cover all the requirement of specific application (for example CM). Ted Lemon: I think these should remain as two separate documents Margaret Wasserman: I do think some of the things mentioned in this document do need to be available, diagram coordination with current MIF API like interface between MIF API and application should be maintained, 21 related would better be a seperate document. Cao: what is the problem for today's connection manager? Carlos: my customer, operator would like to have common user experience for the connection manager. Hui summarized three points need to be considered: architecture alingment with MIF API, Problems of CM, 21 would be a seperate document. 8. draft-deng-mif-api-session-continuity-guide (Suresh, 5 min) Suresh, document has been updated based on the comments from Brian and Pierrick. Would like to see the interests from the floor on this. Margaret, we couldn't ask for adoption without updating the charter. There was a discussion regarding whether happy eyeball can solve this problem. 9. draft-mglt-mif-security-requirements (Daniel, 5 min) presentation finished. Hui mentioned Cao is interested in service discovery in multiple interfaces, please talk to him if interested in it. Meeting adjourned.