?Minutes IETF-73 6lowpan November 18, 2008 Salon F Co-Chairs: Geoff Mulligan and Carsten Bormann Minutes were taken by Anders Brandt and Paul Chilton Meeting called to order at 9am Agenda ------ Admin/Intro - Chairs Routing Requirements - Eunsook Kim Header Compression - Jonathan Hui Bootstrapping/ND optimization - Zach Shelby Whither 6lowpan - Chairs Admin/Intro ----------- Carsten reminded everyone about the "Note Well" IPR. The current charter runs until December 2008. WG work will not be completed by then, but good progress is now being made. Carsten reminded everyone what 6lowpan is about: L2 network: IEEE 802.15.4 Goal: To make a radio link look ALMOST like an IPv6 Link 6LoWPAN Routing Requirement --------------------------- Eunsook presented changes for version -08. The document presents both mesh under and route over. New document was restructured - more details and more examples. 3 new requirements were added and 2 deleted. --- Discussion followed --- Carsten: many requirements were "soft" JP Vasseur: "What is the purpose of the document?", why is this document being written, why not in ROLL Carsten: The 802.15.4 market is well defined. These are routing requirements from 6lowpan and 802.15.4 viewpoint. We do not need to debate this WG Charter item every time. Mark Townsley: The two working groups should cooperate more. If ROLL is going to ignore this document then that is a problem. If it cover 80% of the general routing requirements then ROLL should take those - overlap is good. David Culler: we should move forward with the document. JP Vasseur: is 6lowpan was going to define "mesh under". Geoff Mulligan: we do not know. We are not currently chartered for this. Carsten: should this doc be a WG document. Mark asked if there were alternative documents; are we chartered for this and would David and JP help shape the document. David answered yes. Make recommended the document be made a WG doc. 6lowpan Improved Header Compression ----------------------------------- Jonathan Hui presented -01. Changes to Traffic Class and Flow Label expanded to 2 bits Hop Limit expanded to two bits with discrete hop limit encoding Context identifier extension for global address compression Source address my be compress in stateless or context-based methods Multicast DA now supported via "M" bits. 4 modes supported. --- Discussion followed --- Geoff Mulligan: Hop Limit we are using 2bits to save 8bits and are we sure this will really be the most used formats? Jonathan: Yes and it is not obvious yet. Geoff Mulligan: UDP checksum compression – this is then not UDP. Erik Nordmark: you should consider assigning a new IP protocol transport since it is not exactly UDP David Culler: we should address the efficiency of the compression. If guess too much, risk that most efficient formats will never be of use. Carsten: this was a good document showing what COULD be done, but we do not know what SHOULD be done. Erik: the use of the link local addresses will become quite limited as users start defining the protocols they want to run on the 6lowpan network and we cannot depend on the compression efficiencies gained. 6lowpan bootstrapping and 6lowpan ND optimizations -------------------------------------------------- Zach Shelby presented this new doc based on the 5 previous related ND drafts. This document is MAC agnostic. Nothing is assumed from the link layer. It is also agnostic to router over and mesh under. Defintion of "Link": The subset of neighbors that hear your packet when it is transmitted. This solution is derived from RFC 4861. --- Discussion --- Erik Nordmark: packets received via whiteboard may potentially help header compression. Zach said that whiteboard in every edge router may act as proxies - two pans may form a bigger subnet. Carsten: this document is "big"; thinks that DAD needs further definition. David Culler: we need to clearly state if this is really DHCP or not and if not why this is not DHCP. There was a bit of discussion about if this is just a different form of DHCP. Zach points out that state is in the node not in central server. Question if this is stateful ND? Erik: there is no such thin as "subnet-local address". This may be an issue in 6man. Further discussion surrounding multi-link subnets and TTL problems. Carsten asks if this should be a WG doc? Consensus in room is that it should. Whither 6lowpan --------------- -- Discussion about Architecture doc: Carsten: Should Arch doc hold ND, Routing and use cases David: We had it earlier Mark Townsley: I supoprt having a GOOD arch doc - not at least for the case of future participants Consensus that the arch doc should be revived -- Discussion on Security: Carsten: seems like nobody has security concerns - comments are rare. Daniel: I received a few comments Geoff: I'm also concerned that we are too quite. Mark: Connect with security guys -- Interop: JP: What about interop testing? Geoff: do you suggest that WG do interop testing Agreement from group that 6lowpan should NOT do interop testing but Interop document is within scope. need volunteers. Interop issues should be discussed on list.