Path MTU Discovery WG (pmtud) Monday, 19 March 2006, 17:40--18:40 =================================== CHAIRS: Matt Mathis Matt Zekauskas Scribe: Kevin Lahey These minutes were lightly edited by the Chairs from notes taken by Kevin Lahey. AGENDA: 1. Agenda bashing, milestones review 2. PMTUD method draft update and issue summary 1. Agenda bashing, milestones review Matt Zekauskas opened up the meeting by pointing out that the group was almost done, as all of our milestones have been hit but the MIB. Lars Eggert and Magnus Westerlund, the new Area Directors, said hello to everyone and suggested that folks explain stuff in as much detail as possible, as some of this was new to them. 2. PMTUD method draft update and issue summary John Heffner then gave a presentation on the current state of the draft and on his implementation of the draft in Linux. The code is in the Linux Developers Tree and should appear as part of Linux kernel 2.6.17, probably within two to four months. In response to some comments from Michael Tuexen and Randall Stewart about the SCTP implementation in Linux, John pointed out that his patches provide for some protocol-independent data structures describing the progress of PLPMTUD, and explained that it doesn't yet handle shared state. It was pointed out that Kevin Lahey was working on a NetBSD implementation as well. In the revision from -05 to -06, John and Matt added information about using the new SCTP pad chunk (still being defined) to implement this, fixed some language issues brought up by Gorry Fairhurst, and fixed up the all of the "TODO" items. There remain a few small editorial nits. Then discussion turned to the issue of a MIB. John had a slide with some of the elements that would be useful in a MIB, but as Matt Mathis pointed out, it will be tough to define MIBs for each of the different upper layer protocols that can implement PLPMTUD. He suggested that we could write a template that could be applied to the various ULP MIBs. Then Lars (the AD) asked if we really needed a MIB, and Matt M. explained that it was useful in debugging issues with stuck connections. John explained that it would be very useful for large web servers with many connections, where debugging could otherwise be very difficult. The question then turned to the right forum for the MIB definition -- PMTUD or TSVWG. TSV has a bit more MIB experience. This was not completely resolved. The current plan is to release a new I-D somewhere around April 10, freeze it, at last call, and proceed with the MIB. Once the MIB is written, we can revisit the PMTUD draft to make any last minute tweaks. There was discussion of revising the milestones to reflect this, currently leaving the MIB within PMTUD, with a first draft by next IETF, and completion end-year. The first MIB draft will be a personal submission, and it could be associated with TSV at that time if it is deemed appropriate. Fred Templin finished off the meeting, pointing out that his draft considering PMTUD issues in tunnels, draft-templin-mtuassurance-00.txt, would be discussed in the Softwires working group. With that, the meeting ended.