Draft MEETING MINUTES Ad-Hoc Network Autoconfiguration (autoconf) Meeting : IETF 71 Tuesday March 11th 2008 Time : 1300-1500 Location : Franklin 11/12 Chairs : Thomas Heide Clausen Shubhranshu Singh ========================================================= Note Takers: Charles E. Perkins, Teco Boot ============================================ AGENDA * Administriva o Mailing list: autoconf@ietf.org o Scribe(s) o Blue Sheets o State your name at the microphone o IPR - RFC3979 * Agenda bashing * WG Progress * Mobile Ad hoc Network Architecture http://www.ietf.org/internet-drafts/draft-ietf-autoconf-manetarch-07.txt Currently in state "AD Evaluation :: Revised ID Needed" Recent discussions subsequent last IETF: http://www.ietf.org/mail-archive/web/autoconf/current/msg00924.html The WG needs to resolve this, move document forward. * Address Autoconfiguration for MANET: Terminology and Problem Statement http://www.ietf.org/internet-drafts/draft-ietf-autoconf-statement-04.txt Considerable update subsequent to feedback at last IETF and on list-discussions. * Updates: http://tools.ietf.org/wg/autoconf/draft-templin-autoconf-dhcp-11.txt draft-bernardos-autoconf-solution-space draft-bernardos-autoconf-evaluation-considerations draft-bernardos-manet-autoconf-survey Agenda bashing Architecture document is with the IESG. Status is "revised ID needed". Fred: The reason for delay of architecture document seems stale -- why has it not occurred over the last few months? Jari A.: There has not been very much traffic on the mailing list. I've asked whether Thomas is O.K. with document. Thomas C.: We reached an understanding in San Diego about manet nodes that have manet router interfaces. Narten asks what is the link model and the addressing model. Thomas N.: Requested clarifications for questions already raised on mailing list. see review comments sent to the ML. Thomas C.: Document would benefit from separation into sections answering three main questions. Some of these questions were only settled very shortly before the draft cutoff so could not produce updated document. Problem Statement presentation: Emmanuel B.: It's about autoconfiguration of routers in a manet. - DHCP assumptions are wrong for manet Joe M.: Some people are using DHCP, so DHCP is not _always_ wrong! It would be nice if the slides for the proceedings did not give the wrong impression Thomas N.: If the relay agent is configured with the address of the server, then what is the problem? At this point, I don't even know how to transmit packets on a link? Charlie P.: What happened between [manet] and [autoconf] to make it so hard to understand how to put an IP packet over a link? Joe M.: Strong disagreement with statement about absolute "wrong"ness of DHCP. Dave Th.: How does one replace "broadcast" as needed by DHCP? How does one configure the relay agent? Thomas N.: How does a client communicate with the relay? - NDP/SLAAC: Again, assumptions are "often" wrong in manet. NDP needs a single multicast-enabled link. Thomas N.: There was a call several weeks ago, and a manet link seems to be a "new" kind of link? Charlie P.: Then how is it we have been able to do IP over these things for the last 10 years. Joe M.: Should say "not effective" instead of "wrong". Erik N.: Should determine what are the properties of a link that are preserved in [manet]. Are link-layer multicast properties preserved? Is broadcast semantics preserved? Dave Th.: Architecture document should describe the properties of the links that are being considered. Perhaps define a specific link "type". Then can say how IP can run over the link type. Thomas C.: Have been running revised version of NDP and also some DHCP over the links. All manets are not equal. The architecture document should be careful to not exclude some solutions that use. Emmanuel B.: These mechanisms can work in some manets but in other manets additional mechanisms are needed. Erik N.: There have been in the past attempts to make protocols work over, e.g., non-multicast links and the answers seem to revolve around making the new kinds of link have the properties needed for the protocol. Thomas N.: Maybe there would be a need for some intermediate layer that supplies the missing link semantics for whatever properties are needed (e.g., NBMA links). Dave Th.: agreeing with Erik N. and Thomas N.: what choices are being made for properties to expose to IP? Thomas N.: Is there a plan for describing the link types in the document? Teco B.: DHCP Joe M.: Is the goal of the exercise to delineate the classes of problems that the group is working on? Charlie P.: The properties of the link type will determine the kinds of solutions that are appropriate. Alexander P.: So will there be any attempt to make the properties of the links explicit in the architecture document? Charlie P.: The solutions document authors ought to describe the properties of the links that they rely on for their design. Updates on Autoconf IDs: draft-templin-autoconf-dhcp-10.txt has been updated to ver-11 . This ID was presented in the previous meeting. Carlos very briefly updated status of draft-bernardos-autoconf-solution-space draft-bernardos-autoconf-evaluation-considerations draft-bernardos-manet-autoconf-survey. See presentation material