Monday 3rd December 2007 1740-1950 AUTOCONF WG ============================================== Agenda: o Agenda bashing & Working Group status update - 05 min WG Documents: o Mobile Ad hoc Network Architecture ID Progress - 15 min draft-ietf-autoconf-MANETarch-07 o Problem Statement discussion - 45 min draft-ietf-autoconf-statement-03 Individual Submissions: o Ad-Hoc IP Autoconfiguration Solution Space Analysis - 15 min draft-bernardos-autoconf-solution-space-00 o Evaluation Considerations for IP Autoconfiguration - 15 min Mechanisms in MANETs draft-bernardos-autoconf-evaluation-considerations-01 o Requirements for IP Autoconfiguration Mechanisms in - 10 min Backbone Wireless Mesh Network scenarios draft-bernardos-autoconf-backbone-mesh-reqs-00 o NEMO for Mobile Ad-hoc Networks - 15 min draft-boot-autoconf-nemo4MANET-00.txt Agenda Bashing -------------- Fred Templin (FT): o noted that agenda was not on-line well in advance MANET Architecture ID progress ------------------------------ Shubhranshu Singh (SS): o draft-ietf-autoconf-MANETarch-07 has been sent to the IESG. Further revision to incorporate comments received following meeting. Also to be reviewed by Thomas Narton Jari Arkki (JA): o made first pass, needs to make second pass. Looking good. Problem Statement Discussion: ----------------------------- Thomas Clausen (TC): o We have spent a lot of time on the problem statement document ~2 years and I feel its partly because of dependancies on the architecture document During this discussion I would like to have a goal of defining what we want to configure. o we need to make progress on Problem Statement. Rather than concentrating on text, need to agree on problem. We do not currently seem have common understanding. JA: o WG has been in existence for two years, need progress in months, or weeks. Presentation by Emmanuel Baccelli (EB). Key features from MANETarch document (MANET router, semi-broadcast interface, fuzzy neighbourhood, mobile ad hoc). MANET routers must be able to acquire a unique prefix, over which it has authority. Either subordinate MANET with topological correctness or autonomous MANET. Dave Thaler (DT): o What I think it should say. It should say that a MANET router wanting to acquire a unique prefix. If it only has one address then it does not need a prefix. o if a mobile router has no attached hosts it can just acquire a unique address. EB: o is a /128 a prefix? DT: o You need an address which you can assign to your MANET interface. You also need an address for a loop back interface. Ryuji Wakikawa (RW): o agreed, although addresses != prefixes Joe Macker (JM): o may have different solutions to different cases, does not feel this is captured. Alexandru Petrescu (AP): o is subordinate MANET a new terminology? EB: o yes TC: o indicates if prefix is imposed externally o What is trying to be said here is that a point of attachment imposes an address hierarchy and there is another case which doesn't have an attachment point and therefore doesn't have a hierarchy AP, RW: o needs clarification RW: o This prefix is it the prefix assigned to the MANET interface or something different. And is this prefix different from the network wide prefix. It makes sense to be the MANET local prefix but I think the terminology isn't clear. TC: o MANET Router with hosts needs a delegated prefix, like all routers. An external network may impose a topological requirement. Could be multi-homed, but not shown in figure. o The way I understand the figure in the slide. The blue cloud is full of MANET routers and some have stuff hanging out of them. Some may have prefixes assigned to them which they can then delegate to the stuff hanging off of them. The question is how do these MANET routers get the prefixes which they can then give out to the stuff hanging off of them. o What happens when you have just one router with a single host embedded in it? It should get a prefix for that host as the address on the MANET interface is different. o If this blue cloud is attached someplace to an external network which is imposing a hierarchy then they need to be topology correct. EB: o it is clear who has authority for each prefix. TC: o need topologically correct prefixes when connected to access router. AP: o there's an issue, we need to be clear what "unique, non-overlapping prefixes" mean [editors note, see mailing-list discussions prior to this IETF meeting] EB: o is there disagreement on whether a router without hosts need a prefix? Roland t'Veldt (RV): o if a MANET router has a co-located single host attached, does it need a prefix? TC: o yes. DT: o according to MANET architecture document, yes. o Host don't get access to the MANET interface directly to protect the MANET interface to protect the other MANET interfaces. Charles Perkins (CP): o Right now it's useful for host to have just addresses on the interface and not have prefixes and the extra work require to give prefixes seems like extra work which does not need to be required to be done. o need to relax requirement. Same discussions were held a year ago. Where is the danger? Are we not agreed on this. Teco Boot (TB) o isn't this a transit network? Do we want to autoconfigure transit networks? Aren't these out of scope. EB: o isn't this confusing transit and multi-homing? TC: o as soon as mobile routers have hosts, they are transit networks in this sense. JML o this is also a routing problem. Not clear have good scenarios. TC: o must not a priori exclude this case. DT: o need to discuss multi-homing in problem statement. o Should be allowed but not required. o Need to avoid solutions, but should set bounds. Justin Dean (JD): o a lot of this is about getting addresses to nodes. o Need to consider partitioning, merging, how to return prefixes. o addressing and maintenance are two different aspects. EB: o how much of this is solution? This is in Problem Statement. Gabriel Montgenegro (GM): o There is separate between MANET and non-MANET interfaces to avoid confusion between applications running on the MANET interfaces. Should we say which applications are ok to run on MANET interfaces directly. o I am concerned about running multiple interfaces on very small 8-bit type processors. DT o This is a MANET arch document issue and it is included in there already to allow applications to specify if they can be run on a MANET interface, right now its mostly just MANET routing protocols but it is not mandated to be all. EB: o input to enable substantial rewrite of document. JM: o still not clear if group is clear on what it wants to achieve. o What more than addressing assignments? o Early on, there was a lot of discussion regarding service discovery, etc.? TC: o not doing things that aren't address related. TC: o Now, are there anyone present who have any subject/concern/issue with the PS document that they want to discuss, but which hasn't been brought up yet? If so, now is the time... AP: o We are discussion problem statement and we have only seen goals. Stateful/stateless which MANET protocol are we working with. DHCP or not? o I would also like to see possible mobility files to show which protocols break and in which cases. TC: o ND/DHCP/etc. are solution space, not problem statement. AP: o when do we know? TC: o believes the the group is not at the point where it can claim consensus on which solutions to pursue just yet. ??? o Suggest adding section on why can't just use existing solutions. EB: o Section 4, but not as clear as it should be. RW: o need sections to discuss specific protocols. EB: o currently more abstract, needs improvement. JM: o Need to add multi-homing specifics, currently left up the reader to deduce. EB: o there is a section on multi-homing. RV: o want clear requirements list. CP: o repeat comment that prefixes may not be required by each router? o I don't understand why a normal application can't just run on a MANET interface TC: o see architecture document and documents by Dave Thaler, enumerating what problems can arise if they do. ??? o Have been comments on AP: o have goals. Want to be able to compare solution with requirements. TC: o need clearer description of two cases, of when a prefix is needed, to clarify what "unique prefix" means (two possible meanings, only one meant), transmit case, what are objectives? o Asked Emmanuel Baccelli to summarise to mailing list. Aim is document to IESG by next IETF meeting. JM: o do we need to meet all goals? TC: o not discussed exhaustively. (As individual, he wants solution that covers everything, but modular to allow not include what don't need.) draft-bernados-autoconf-solution-space-00. ------------------------------------------ Comments invited on mailing list. draft-bernados-autoconf-evaluation-considerations-01. ----------------------------------------------------- Comments invited on mailing list. draft-bernados-autoconf-backbone-mesh-reqs-00. ---------------------------------------------- Comments invited on mailing list. draft-boot-autoconf-nemo4MANET-00. Comments invited on mailing list. ==== THANKS to our: Jabber Scribe: Dave Thaler Minute Takers: Christoper Dearlove Justin Dean