IETF 94 - Homenet Tuesday, November 3, 2015 0900-1130 Morning Session I Room 502 Note taker - Stuart Cheshire Jabber relay - Dave Thaler --- 1. WG Status Update - Chairs (10m) Updated Drafts: - draft-ietf-homenet-prefix-assignment-08 [in RFC Editor’s Queue] - draft-ietf-homenet-dncp-11 [Approved!] - draft-ietf-homenet-hncp-09 [IETF LC closed 23 October, IESG Telechat 19 Nov.] - draft-ietf-homenet-front-end-naming-delegation-04 - draft-ietf-homenet-naming-architecture-dhc-options-03 - draft-ietf-homenet-hybrid-proxy-zeroconf-02 - draft-geng-homenet-mpvd-use-cases-02 Question: How can an experimental protocol be mandatory? Mark Townsley: If you want to participate in the experiment then this part is mandatory. --- 2. Autoconfiguration - DNCP / HNCP IESG review update - Steven Barth (15m) - draft-stenberg-shsp-00 - Markus Stenberg (10m) Lee Howard: I support this direction. Sadly the industry right now seems to be moving in the direction of “The Compuserve of Things” --- - draft-barth-homenet-wifi-roaming-00 - Steven Barth (15m) Lorenzo Colitti: This seems very complicated. Why not just do bridging? And then multicast would work too. Steven Barth: Because this would be more efficient. David Lamparter: This would break direct client-to-client communication (802.11z). Steven Barth: I agree. Mikael Abrahamsson: You got this working, right? How much loss of connectivity was there during roaming events? Steven Barth: A second or two. Mikael Abrahamsson: Would this work with Lorenzo’s proposal for a /64 per host? That would simplify things. You would no longer need a DAD proxy. Steven Barth: You could run out of prefixes very quickly. Pierre Pfister: We tried a /64 per host and had it working at Bits ’n’ Bites at the last IETF. Lorenzo Colitti: Some issues about roaming in the home: 1. How quickly do you notice that you’ve moved? Android will notice in about 3 seconds. But then it takes a few seconds to reconnect. 2. How much does it cost to move? You have to kill and reopen TCP connections. TCP Fast Open can help. Steven Barth: Roaming between LTE and Wi-Fi has the same issues. MPTCP and QUIC can help. David Lamparter: VoIP needs roaming times in milliseconds, not seconds. IEEE 802 is aiming to achieve this. Michael Richardson: The proposal may look ugly, but link-layer bridging looks ugly too if you understand how it works. I also want roaming to work from Wi-Fi to Ethernet, not just one Wi-Fi AP to another Wi-Fi AP. Mark Townsley: If all upper layers (TCP and above) already handled client IP addresses changing gracefully, we’d have nothing to do here. But if link-layer doesn’t handle this, and TCP and above don’t handle this, then the Internet layer needs to play a role in handling this. Lorenzo Colitti: You can’t get roaming times in milliseconds without support at the link-layer. * Michael Richardson will work with Steven Barth on updating the draft. --- 3. Naming Architecture and Service Discovery - draft-ietf-homenet-hybrid-proxy-zeroconf-02 - Markus Stenberg (10m) Stuart Cheshire: I think this is a good idea that solves a useful problem. Michael Richardson: This also needs a loopback address. Markus Stenberg: We need some (dummy) interface that’s always up and available. Ray Bellis: I think this is a problem that needs to be solved. But doing multiple queries may not be the best answer. In an enterprise logical grouping is useful. In a home a single flat namespace is generally more useful. Which hosts and/or services are mirrored into the public DNS? Markus Stenberg: This is an implementation detail. Steven Barth: Host names from DHCP are also useful, but DHCP servers currently have no protocol for communicating between themselves to check uniqueness of the host names their clients claim to have. Ted Lemon: Perhaps DHCIDs can help. Yoshiaki Tominaga: What if there are two printers? Stuart Cheshire: Printers have factory default names, for example, “Epson WF816”. If I install a second printer of the same kind it will detect that the default name is already in use and rename itself “Epson WF816 (2)”. --- - draft-ietf-homenet-front-end-naming-delegation-04 - Daniel Migault (15m) + draft-ietf-homenet-naming-architecture-dhc-options-02 Ted Lemon: Why would the Homenet Naming Authority not be on the CPE? Mark Andrews: Being able to run the Homenet Naming Authority on a “real” host with a reliable file system that’s backed up can be beneficial. Ray Bellis: Use of the DHCP options assumes the Homenet Naming Authority is colocated with the DHCP server. Steven Barth: If Homenet Naming Authority functionality is optional then user may end up with a network with no Homenet Naming Authority available. Stuart Cheshire: Other IETF docs take the approach of making functionality mandatory to implement but not mandatory to use. That way every Homenet- compliant CPE must offer Homenet Naming Authority, and then the customer is free to decide which device they’d like to use to provide that functionality. Michael Richardson: Not requiring link names and router names embedded in host names is important. Douglas Otis: If one device has multiple interfaces we need to know if it’s one device with multiple interfaces or multiple devices with one interface each. Dave Thaler: If I have a domain name I’ve paid for, can I use that for my hostname (via DNS Update) from outside the home too? --- 4. MIF + Homenet - draft-geng-homenet-mpvd-use-cases-02 - Liang Geng (15m) - draft-stenberg-mif-mpvd-dns-00 - Steven Barth (15m) Ted Lemon: I was concerned about the examples. A VPN that’s terminated at the network edge is not very secure. Proprietary boxes with a walled garden are not an outcome we’re trying to encourage. Ole Trøan: My home is a provisioning domain (administrative boundary). I don’t want some external entity pushing configuration information into it. Mikael Abrahamsson: Other people might actually want that. Juliusz Chroboczek: The fact that there's an additional signature seems to imply that the security considerations for this information are different. Could you please explain why this is the case? Markus Stenberg: DHCP is hop-by-hop. The MIF provisioning is end-to-end from ISP. Juliusz Chroboczek: Thanks. Michael Richardson: This won’t work because the people who build walled gardens don’t know enough to configure it that way. Markus Stenberg: We need to provide guidance to walled garden operators. Lee Howard: It seems like an awful lot needs to happen to make this work. Erik Kline: Will this survive in reality? Steven Barth: It’s no harder than the existing reverse-mapping DNS records. Mark Townsley: The Homenet and MIF chairs will be getting together to discuss this. ---