breakout summary 1

[Jun-ichiro itojun Hagino <itojun@iijlab.net>]

[not an official minutes]

homework
	rob: dnssec and dnsalg analysis


breakout actions

ISP:
added additional work areas (and volunteers)
	wireless
	broadband ethernet
	infrastructure services (overlap area
		data centers
outline clarifications
	add multicast
	add multi-homing/managed access
	security
	traffic engineering
dsl detail
	L2 discussion not necessary
	other issues send to list
agreement that the draft represents the majority of ISP networks
IIJ presentation on IPv6 ISP in Japan
	provides input for the analysis draft

wireless: keep it, or move it?  still open question


3G:
(missed a slide)
IMS UE connecting to a IPv4 SIP node
	not as general as normal Ipv6 only node scenario - SIP-ALG vs DNS-ALG
	two cases exist
		SIP ALG and media flow translation in the same box
		SIP ALG and media flow translation in the separate box
	next steps
		mroe text to clear up the analysis
		protocol between SIP ALG - media flow translation
GRPS IPv6 only case
	a long discussion of the problems of NAT-PT
		DNS ALG - breaks DNSSEC
		scalability issues - how to use multiple NATs
	NAT64 vs NAT-PT
		having DNS ALG in DNS server in the network
		having A record support in the end-node itself
			DNS-ALG in end-node
		trust models
			trusting operator DNS server
			not trusting operator infrasturcutre
		hybrid model

huitema: trust model change is bad - it's not the main topic of 3G, common problem
mar: NAT-PT issue applies to other uses? - unsure.
mar: scalability? - no real diffs.
rob: DNSSEC issue needs to be digged further
itojun: dnsext or dnsop or v6ops?
rob: not sure, anyways rob will volunteer (before atlanta)
alain: need to go quick, IPv6 only device will start to appear


unmanaged:
unmanaged team report
	get organized
		list of issues
		list of other works
	the name resolution issue
	multiple link issue
list of issues
	different topologies (routers, multiple routers)
		multihousing unit, shared subnet, multiple subnets
		shared wireless (single/multiple isp, security)
		fixed ethernet (impacts, naming, discovery)
	prefix delegation
	use of tunnel broker for case A - dualstack host, v4 gateway, v4 isp
	monitoring requirements
		how do you debug your network - some discussion
	SIIT issue: only layer 3, support for port mapping, v4 to v6 support?
	registration of new appliances, notification, device capability, security (application vs network)
		mar: what is related to IPv6 deployment? - that was the discussion
	security: is the inside really safer?
	what support do we need for legacy ipv4
		ipv4 only, local connectivity to local ipv6, remote ipv6 (legacy apps vs new apps)
	solutions for name resolution -> LLMNR, DDNS, etc
	mobility, roaming -> guest in the house, call back home

work for another team
	ad hoc (car)
	personal area network
	mobile network 

mar: 4 teams won't cover every type of networks.  it's not possible to cover all possible cases

name resolution
from A to B	B v4 only		B dual		B v6 only
		---			---		---
A v4 only	v4(B)			v4(B)		some translation
A dual		v4(B)			v4/v6(B)	v6(B)
A v6 only	some translation	v6(B)		v6(B)

problems with NAT-PT
	we don't want to use ISP NAT-PT box
		A dual, B v4 only
		A dual, B v6 only

recommendations for naming
	IPv6 host initially be dual stack
		ipv6 host will also look for A record
		solves the "literal URL" issue
	if not dual stack, use "local SIIT" (BIA)
		need to reserve of configure a SIIT prefix
		or just forego interoperability with ipv4
	leave the AAAA requests alone
		sincce the host will look for A if needed

bound: why are you digging this far
huitema: we're already doing analysis document, not scoping document
thomas: how can you talk to IPv4?
huitema: i'm talking about naming.  header translation in network somewhere.

result of dual-stack or SIIT
	v4 only -> v6 only, no solution

more naming recommendation
	ipv4 only host need some translation
		need a DNS ALG
	but translating A requeest is harmful to dual stack hosts
	suggestions
		use two different DNS services, based on protocol type, port number or names
		use a special address range for translated addresses
			class E?

itojun: which node needs changing?
huitema: dualstack, or ipv6only

more naming stuff
configuration
	both dhcp option and reserved address work
dynamic DNS - stateless, can lose database
acquiring autoconfigured addresses
	gateway receive request for local name
	gateway issues LLMNR request, cache the result
	returns AAAA reply
advantage: stateless

rob: conflicts with dnsext recommendation on LLMNR (TTL and such)
hain: is it the name leak problem? - yes

multiple links
some medias are hard to bridge
	bluetooth, ieee1394, possibly power line
request to support two topologies
	star
		every link goes to the router
	mesh
		routers connected to the home network

connectivity issues
two options for mesh topology
	multilink subnet (single /64)
		proxy arp
		need a detailed specification
	configured subnet prefixes
		require multiple /64 prefixes
		require a prefix allocation specification
star topology
	simple solutions.  e.g. allocation of subnet # by the gateway + use of classic RS/RA

huitema: no solution for mesh topology for now. - hole.

naming issue in star topology
configuration
	both dhcp and reserved address work
LLMNR for resolving AAAA
	gateway must repeat the query on multiple links

naming issue in mesh topology
configuring the dns server
	the reserved addr approach just works
	dhcp requires a set of dhcp proxies
options for using LLMNR
	LLMNR cache + relays at routers
	or, "right scope" multicast
		subnet scope if multilink subnet
		site local scope if configured routers
multicast issue appear in LLMNR, but also duplicate address detection

thaler: dynamic dns? - (lost answer)
mar: how do you discover addr?
rob: dynamic update needs to be explored
?: scope really is a domain?  is it really a DNS domain/scope?
templin: meshed network case has a lot of overwrap with adhoc network space