IETF 104 - Homenet Agenda Prague, Czechia Tuesday, March 26, 2019 9:00-11:00 Tuesday Morning session I Grand Ballroom ============================================================================ Administrivia and Status Update: Chair slides (10 minutes) Chairs: Stephen Farrell, Barbara Stark - Blue Sheets - circulating - Note takers - Evan Hunt, Brian Haberman - Jabber relay - Mikael Abrahamsson - Thank you, goodbye, hello and welcome (AD changes) - Status of drafts Daniel Migault not present, so status moved till later. ============================================================ Simple Naming (including any Hackathon readout), joining IoT edge routers to Homenets, and what are we actually producing anyway? https://tools.ietf.org/html/draft-sctl-dnssd-unicast-autoconfig-00 and other drafts (Ted Lemon, 20 min) ----------------------------------------------------- Homenet Marketing document (https://tools.ietf.org/id/draft-lemon-homenet-review-00.txt) - market analysis comparing features among CPE products - Focused on multi-router home networks - Needs review and feedback ----------------------------------------------------- Homenet in OpenWRT - progress made on code, not much on the document ----------------------------------------------------- Simple Naming - Have home.arpa - No way to get info into authoritative server - discussion of missing/unspecified features in hncp (link naming...) Robin Wilton: Regarding link naming: to what extent is this dependent on vendor behavior? Ted: we need to specify the correct behavior Ted continues to present slides. - per-link authority requires choosing dnssd proxy for each link, which would be done by HNCP; the HNCP extension is not done or specified - also requires auth server for home.arpa to delegate to each link; not solved for more than one router. - reverse mapping is not hard but not yet implemented; need to be able to advertise reverse name mapping registration protocol. This is not a high priority. - name resolution - local names done except for authority setup and delegation; global names done as well via discovery proxy - in stateful configurations, secondary routers need to be able to forward home.arpa queries to the authoritative router Stephen Farrell: There will be difficulty with DoH. Ted: Prefer using DoT with known external DNS servers. - dns push is required to get feature parity with mDNS. This is done. Bernie Volz : how do security associations get established? Ted: relying heavily on HNCP. TLS is hard. Ted continues to present slides. - provisioning domains: not done, need the PvD RA option first Juliusz Chroboczek: How will this work with multiple ISPs? Ted: Ideally we use provisioning domains if available, but not yet widely supported. So we round robin among source addresses until one works. Juliuz: What about if two ISPs provide nonequivalent name servers? Ted: Host is round-robining amongst different DNS servers. Only alternative is to use NAT which would be a bad idea. Mikael Abrahamsson: This is not a solved problem. Should we write a v6ops problem statement document? Ted: Reluctant to deal with v6ops, perhaps a short-lived working group would be better. Éric Vyncke: Ted: Could support multihoming only if support for PvD, and be effectively single-homed if not. Christian Franke: Round robin at the host performs badly if there are resolvers that don't work. Ted continues to present slides. - Service Registration Protocol: Work still needs to be done. SRP proxy is incomplete. might get support in BIND 9, could also be done in mDNSresponder - Homenet and IoT + two things necessary: isolation and routing + isolation: IoT devices must be reachable on wifi but not by all hosts. Need to be able to isolate nodes on the same SSID, is this possible? Jacques Latour: Discusses secure iot gateway project based on MUD Stuart Cheshire: asked for more info from Jacques who will provide a link Ted continues to present slides. - routing: 6lo is working on a "routing" proposal to have a single backbone as a bridge for iot devices. Ted is skeptical about scalability. could HNCP be used to make it work better? - Next steps (Ted's view): Major vendors may not adopt but homenet can happen in openwrt. If it does, a lot of cheap and super-expensive routers could wind up supporting it. It could then spread. If it doesn't happen, homenet probably fizzles. Juliusz: Problem isn't that it doesn't work in openwrt, but that it isn't enabled by default. May be politically difficult to get it turned on by default in openwrt but easy to fork openwrt and produce images. Mikael: more of a custom profile than a "fork" Ted: Most things work, except for naming, so that is current focus ============================================================================== Status of drafts from Daniel Migault: - two expired drafts: 1. how to outsource your desired public name to a DNS provider - this should be revived (front end naming architecture) 2. how to configure. Not so sure about whether 2 is needed now. =============================================================================== Where do we go from here? (Chairs, 30 min) - Lack of feedback from mailing list on future of WG - Open question as to whether a re-charter would help. - Will a multi-router home network be more prevalent than a home network with a single router and lots of IoT devices? - Is it a marketing/communication problem? - lengthy discussion of use cases and problems that are not being well communicated, e.g., lack of support for smooth handoff between broadcast domains - should we re-charter? are we failing to discuss these things because we believe they're out of scope? - What could be done that would get people more active on the mailing list? - Hum on re-chartering - room prefers to attempt re-chartering; needs to be confirmed on the mailing list. - Barbara will post issues from current charter github site to list to try to get things started. Meeting adjourned 10:48