Softwire @ IETF 81

Stuart Cheshire <cheshire@apple.com>
* Wizard Without Portfolio, Apple Inc.
* www.stuartcheshire.org

0900 Chairs, Welcome notes and agenda bashing

Alain: 6rd Operational document is due by August 2011

Mark Townsley: Work on this is underway, with operator input, and interacts with work in homenet and v6ops

0910 Frank Brockners, Gateway Initiated Dual-Stack Lite Deployment

Frank Brockners presented document status update

0915 Roberta Maglione, RADIUS Extensions for Dual-Stack Lite

Version-04 submitted this morning

No comments from the room

Alain: These changes are significant, so we will do a short WGLC

0920 Yiu Lee, Deployment Considerations for Dual-Stack Lite

No comments from the room

Alain asked for consensus on adopting as WG doc: Many hands raised for adoption, none against

0926 Mingwei Xu, Softwire Mesh Multicast

Alain asked room for questions. There were no questions.

Alain: This is in our charter.

Alain asked for consensus on adopting as WG doc: 13 hands raised for adoption, none against

0933 Yiu Lee, Multicast Extensions to DS-Lite

Behcet Sarikaya: It's misleading to describe this as "stateless"

0940 Behcet Sarikaya, Multicast Support for Dual Stack Lite

[...]

Xiaohong Deng: Which ISP networks would use this?

Behcet: When many clients join the same group and you're concerned about lots of packet duplication

[...]

Alain: We have three ways here to do multicast with DS-Lite. Do we need all three, or just one?

Tina Zou: One

Rajiv Asati: We need a comparison of the methods

Alain asked room for who supports each of the three techniques:
Yiu Lee's approach: 10
Mcast inside tunnel: 1
L2 optimization: 0

Alain: We'll have one-week working group

0957 Behcet Sarikaya, Multicast Support for 6rd

Mark Townsley: If we can't make this fully stateless, then why not just use existing AMT protocols

1000 Satoru Matsushima, Motivations for Stateless IPv4 over IPv6 Migration Solutions

Alain asked for comments from the room. There was none.

1005 Alain gave presentation outlining differences between CGN and Port-Range-Router scenarios

Mark Townsley: The presentation does not accurately reflect what is being proposed by the proponents of stateless operation

Rajiv Asati: Why does NAT binding include customer ID as well as Internal IP address and port?

Alain this is specified in DSLite documents

Satoru Matsushima: No state

Ralph Droms: Should move on to Dave Thaler's presentation

Mark Townsley: Don't put state in the middle of the network

1030 Dave Thaler

Shared IPv4 address solutions assume that "no kind of end host will be part of the shared IPv4 address system."

When such assumptions are violated (as they will be) the failure modes must be considered.

There are two separate issues being debated:
static vs. dynamic port allocation, 
an stateless vs. stateful port allocation

e.g. static port allocation does not have to imply stateless;
static ports can be allocated using a "stateful" protocol too

Mark Townsley: Should a document document the failure modes?

Dave Thaler: Yes

Mark Townsley: There may be two separate issues, but in reality a stateless mechanism requires static port allocation.

Hui Deng: If this work is to be done, it should be done in some other working group outside softwire.

Alain: Who is in favour of adopting this document as WG document?

Lot of hands in favour, none opposed.

1055 Tetsuya Murakami, IPv4 Residual Deployment on IPv6 infrastructure - protocol specification

Alain: Is static allocation of fixed-size port sets adequate?

Mark Townsley: Operators I've talked with don't have trouble picking a port set size that suits them

Dan Wing: Keep document simple. Just do "prefix-based" The "port sets with heads" and "modulo" are too complicated.

Tetsuya Murakami: The scattered static allocations are required to facilitate IGDv1 clients that try to grab ports sequentially

Dave Thaler: For security reasons we want the algorithmically-allocated static port set to look as much like random numbers as possible, to limit ability of attackers to guess which port a client will use next.

1115 Peng Wu
Public IPv4 over Access IPv6 Network
DHCPv4 Behavior over IP-IP tunnel

Ralph Droms: I think "DHCPv4 Behavior over tunnel" document should be in DHC working group

Alain: Who supports adopting "DS Lite without NAT" as WG document?

For: 8
Against: 0

Alain: Who supports adopting "DHCPv4 Behavior over tunnel" as WG document?

For: 9
Against: 0

1125 Gang Chen, 4via6 Stateless Translation
Xing Li: We may also need to consider DIVI as double translation technology.

-- END