IETF 82
Homenet Working Group
Tuesday November 15th 2011

Working Group Minutes taken by Chris Griffiths and Alejandro Acosta

09:04 - Chair Introductions

Presented "note well"
Presented "Administrativia"
Presented "Agenda"

09:09 - Ray presented an Interim update  
   
09:10 - draft-chown-homenet-arch-01

Tim Chown Presenting

Slide 4 - Develop homenet version 1

Slide 5 - Practical Example

Make useful HOMENET recommendations, automatic home networking items.

Slide 6 - ULA addresses

open questions whether we should use ULA, not NAT.

Slide 7 - Topologies

Basic network architectures - RFC 6204 - 

# Pete Resnick - Please speed this presentation up a bit.

Slide 9 - Topology Considerations

Maybe affected by practical issues - e.g. chaining of devices.

# Jari Arkko now presenting

Slide 12 - Dual-stack homenets IPv6-only in the future

# Erik Nordmark:  What does this statement mean?  Any topology that we
# sell should support IPv6 and IPv4 and NATs.  Are you doing things
# uniformorally for IPv4 and IPv6.  If you cannot break IPv4 you cannot
# move.

# Lorenzo Colletti:  If IPv4 is broken, someone will fixit.  

# Ralph Droms:  Are we talking about unmanaged vs managed topology?  The
# features that we are talking about.

# Mark Townsley:  We handle home networks and boundaries.  Self-organizing 

# Dave Thaler:  make IPv4 and IPv6 topologies congruent. treat them as
# ships in the night, but make them follow the same shipping lanes.

# Erik Nordmark:  Speaking

# Tom Herbst:  Speaking

# Brian Carpenter:  Speaking

Slide 14 - Prefix Assignments

Multiple usable prefixes.  Prefixes should be stable across reboots.
This was discussed on the mailing list.  Should be conservative on
number of prefixes.

Slide 15 - Discover Borders

Homenet:ISP, Private:Guest, Route:Bridge

# Peter L:  Box finds borders, if you make it generic, you don't need to
# worry about it.

# Mark Townsley:  Put a architectural constructs around it.

# Lorenzo Colletti:  Some sort of glue in the cloud using one connection
# or partitions.

# Jari Arrko:  One border, anything more complicated will be too
# complicated.

Slide 16 - Other Principles

Conservative approach.  Routing protocols, Multi-homing, avoid making
future renumbering harder.

Slide 17 - The Architecture

Slide 18 - Architecture (1)

Support multiple subnets and routers, use link-state routing protocols
(eg. OSPF), LLNs, VM.

Slide 19 - Architecture (2)

Internal stable and efficient prefix assignment, simple security, local
DNS servers and cross-subnet mDNS.

# Dave Thaler: That should be /4?  It is on the next slide.

Slide 20 - Open Issues

Complete arbitrary topologies, discovering borders, ULA needed?,
discover and naming across subnets.

# Kerry Lynn?:  I don't think service discovery, naming, security policy
# are necessarily L3.

10:11 - Prefix Allocation

Ole Troan presents an overview of the possible solutions

Slide 2 - Problem

Slide 3 - Presentation of multi-home topology

Slide 4 - Possible Solutions (Interim)

Slide 5 - Consider - Flooding vs Request/Reply Jari

# Jari Arrko:  Quick observation on diagram - multiple prefixes on
# multiple interfaces.

# Ted Lemon:  Are you aware of prefix allocated?  Source would be edge
# router, not ISP.

# Lorenzo Colletti: On DHCP, flooding vs protocol.  In DHCP you don't
# have a way to validate?  Only way to validate is person who gave you
# allocation.  Reconfigure.  Should we choose pull model vs push model.

# Ted Lemon:  Should this be DHCP?  If so this might require additonal work.

??:?? - draft-chakrabarti-homenet-prefix-alloc-01 

Erik Nordmark

Slide 2 -

Goals Look at existing IPv4 home networks with multiple NATs.  How can
we add IPv6 to those without any IPv6 NATs.

Slide 3 -

Why multiple customer Routers today IPS provides a box, perhaps ethernet
port, no WiFi.  That box does NAT, daisy chain.

Slide 4 - State of current IPv4 home routers Dedicated uplink port.

Slide 5 -

Core of proposal Just use DHCPv6 Prefix Delegation Start PD server.
Carve up the delegated /N

Slide 6 -

Recommended default k=3 Allows for 7 downlink routers Allows for 3
levels of hierarchy from a /56 This is not for grandma...

Slide 7 - Prefix Stability

Slide 8 - ULA

Slide 9 - Multi-homing

Slide 10 - Next Steps

# Ole Troan:  Ok as long as you are foward compatible.

# Ted Lemon:  Reluctant to work on this for WG.  This will be the way they
# solve the problem instead of future state items.

# Lorenzo Colletti:  Pick one and stick with it.

10:36 - draft-baker-homenet-prefix-assignment-00

Fred Baker presenting

Slide 1 - Allocating Prefixes Proposal

Slide 3 - I suggest we assume distributed control and DHCP/DHCPv6

Slide 4 - Typical configurations

Slide 5 - Simplest Case:  Only one router

Slide 6 - Next-simplest case - Tree Network

Slide 7 - Multihoming

Slide 8 - Where distributed allocation gets hairy Not a simple tree

Slide 9 - Minimizing probability of simultaneous allocation.

Slide 10

# Peter Lothberg: Template has IETF on it.

# Jari Arrko:  I like this better than previous one.

# Kerry Lynn:  Tussle Centralized vs Distributed?  Zeroconf benefits.

# Pascale Thubert:  Topology changes.  We have to worry about renumbering.
# Do nothing at Layer2, do everything at layer 3.

10:49 - draft-arkko-homenet-prefix-assignment-01

Jari Arrko Presenting

Slide 1 - Protocols for Home Networking

Slide 4 - Requirements

Slide 5 - OSPFv3 Prefix Assigment Extentions

# Jari:  Draft is reasonably complete.

# Samita Chakrabarti?:  Are you also doing this in IPv4?

# Peter Lothberg:  I want to know the quality of the network.

11:00 - Mark Townsley introduces Routing

11:01 - draft-acee-ospf-ospv3-autoconfig-00

Jari Arkko Presenting

Slide 2 - History/Requirements

Slide 3 - OSPFv3 Defaults

Slide 4 - OSPFv3 Router-ID

# Lorenzo Colletti:  I think you can use detection.  I do like this.

# Dave Thaler:  Potential insentive problem.  Simple term, does this
# provide homogeneity.

11:11 - draft-howard-up-pio-00

Lee Howard presenting without slides.

How to get out of the network.  We already have RA that provide default
out.  We have the information to find next hop.  How far are we from
next hop.  This is an overloading of the RA protocol.  Looking for
consensus.

# Dave Thaler:  RFC4191 - 6MAN working group, and decided that RA not a
# routing protocol.

# Andrew: Stuff is still relevant, hop count is not sufficient.

# Eric Nordmark: We don't need something new here.

# Dave Thaler:  WG decided not to do this because too much size for RA.

11:20 - draft-vyncke-advanced-ipv6-security-03

Eric Vyncke presenting

Slide 1 - Advanced Security Firewall Middlebox?  Use data from cloud?

Slide 2 - Why is this important

IPv6 & Homenet OS updates, trusted vs untrusted networks.

Slide 3

Opening the can of worms NAT is useless for security most botnet are
behind NAT Allowing PCP or UPnP

Slide 4 - Default Security Policy

# Steven Farrell: ?

Slide 5 - More on Paranoid Openness

# Jari Arkko: Can we trust?

# Stewart Cheshire:  I would buy an off the shelf item to protect the
# network.

# Ted Lemon:  If we had secure certs, actual guy I am talking to

11:33 - draft-kitamura-ipv6-auto-name-00 

Hiroshi Kitamura presenting 

Slide 2 - Index

Slide 3 - Introduction

Slide 4 - Auto name example

Slide 7 - Contributuon in regular resolving

Slide 8 - Contribution in regular resolving name --> address (2/2)

Slide 9 - Contribution in reverse resolving address --> name (1/3)

Slide 10 - Contribution in reverse resolving address--> name (2/3)

Slide 11 - Contribution in reverse resolving address--> name (3/3)

Slide 12 - Deployed notions and functions used in auto names

Slide 13 - Regular and Reverse mapping

Slide 14 - Target IPv6 Addresses of Auto Names

Slide 15 - Design (skipped several slides)

Slide 19 - Site dependent Mapping tables (for collision avoidance)

# Stuart Cheshire - This is not user friendly.

--

The presentation on draft-haddad-homenet-gateway-visibility-00
was skipped due to lack of time

11:40 Session ended