2009-07-01
PRESENT
Marcelo Bagnulo
Gonzalo Camarillo
Stuart Cheshire
Lars Eggert (IESG Liaison to the IAB)
Aaron Falk (IRTF Chair)
Vijay Gill
John Klensin
Olaf Kolkman (IAB Chair)
Andy Malis
Danny McPherson
Jon Peterson
Dow Street (IAB Executive Director)
Dave Thaler
APOLOGIES:
Sandy Ginoza (RFC Editor Liaison)
Russ Housley (IETF Chair)
Gregory Lebovitz
David Oran
Lynn St. Amour (ISOC Liaison)
ADDITIONAL PARTICIPANTS:
Wes Eddy
Two business items were added to the beginning of the techchat. Also, Stuart reported that he and his colleagues at Apple have found cases where IPv6 provides a natural way to solve specific problems that Apple engineering teams are working on. He found this to be encouraging, because it provides examples of cases where there’s a clear reason to use IPv6 to solve a real problem, rather than just using IPv6 because everyone says you should. Stuart requested time on the IAB Sunday agenda in Stockholm to describe these use cases to the IAB.
No meeting minutes were reviewed during this call.
The IAB had previously discussed a possible formal liaison relation with the Messaging Anti-Abuse Working Group (MAAWG). There has been some interim correspondence on the the matter, and during this meeting the IAB formally approved the liaison and appointed Barry Leiba to the role. Barry is a former IAB member with expertise in Internet mail; he had also previously suggested such a liaison relation with the MAAWG.
Due to vacation schedules, and since there were no urgent items needing (voice) discussion in the coming weeks, the board decided not to hold telechats on 7/8 and 7/22. The next face-to-face meeting will be in Stockholm on 7/26. Any remaining IAB business will be conducted via the mailing list.
Wes Eddy is a co-chair of the IRTF Internet Congestion Control Research Group (ICCRG) and the IETF TCP Maintenance and Minor Extensions (TCPM). At this techchat, Wes led the IAB in a discussion of TCP and congestion control, focusing on recent work in the IETF and IRTF, and on congestion control approaches emerging outside the IETF.
Wes began with a background on current traffic characteristics, including type, locality, and the impact of network topology. He then walked through various user complaints and provider issues. There was a discussion of the youtube and Bit Torrent approaches, as well as the new Adobe flash protocol. It was pointed out that content distribution networks significantly alter the manner by which load is spread.
The group then considered the following talking points, which the IAB had raised as part of preparing for the discussion:
– what layer violations or other modifications are being implemented as a result of the TCP’s inadequacies?
– are these new or proposed behaviors widely deployed, and are they dangerous, adequate, and/or worth standardizing?
– are there significant proposals out there that are not getting traction in the IETF (but should be)?
– is there IETF work or other coordination that is needed to make progress?
Wes noted that link-layer technologies have changed much faster than end-to-end congestion control schemes, and that few schemes support the wide range of network conditions that exist today. He also drew attention to the time and effort required to move a new standard through the IETF, as compared to the rapid deployment goals of many implementers and application developers. The group discussed how new models get deployed, often at the application layer, and the move of some technologies into the OS over time. It was agreed that it would be good to try and get some Adobe and Linux developer participation into the IETF. Jon Peterson noted that in the case of Bit Torrent and Comcast, the motivation to standardize through the IETF came largely from political forces. In that situation, the IETF was agile enough to respond favorably (i.e., via LEDBAT).
There was some discussion of a TCP guidelines document that would be similar to the UDP guidelines document. Such a draft could cover areas like “how many concurrent TCP sessions is it appropriate for a single application to open?”. This question is especially relevant in the context of Carrier Grade NATs. The group also discussed the large number of TCP variants and extensions, and that some summary guidance in this area would help implementers decide which extensions should be implemented and enabled.
The meeting closed out with some consideration of moving beyond TCP-friendly as the benchmark for congestion control performance. This has been a topic of recent discussion within the IETF and IRTF.