Draft Minutes IRTF Routing RG Meeting 10 november 2006 Overview These minutes are based on notes taken by Tom Taylor during the meeting] A full meeting was held that covered several topic: - There was a discussion of the two RRG drafts: http://www.ietf.org/internet-drafts/draft-irtf-routing-history-03.txt http://www.ietf.org/internet-drafts/draft-irtf-routing-reqs-06.txt A straw poll was taken during the meeting on whether people had read the documents and whether they thought the documents should be advanced to the IRSG as IRTF RFCs. A number of people had read the document and indicated that they should be advanced. Aaron Falk, the IRSG chair explained the relatively new process in the IRSG which requires that documents advance by Research groups receive substantive review in the IRSg before being sent to the RFC editor for publication. The IRTF process, documented in a lapsed draft: http://www.watersprings.org/pub/id/draft-irtf-rfcs-00.txt include a role for a document shepherd and is currently using the IEFT shepherding guidelines as a rough guideline: http://www.ietf.org/internet-drafts/draft-ietf-proto-wgchair-doc-shepherding-08.txt Since the chair of the RRG is one of the authors and editors of the drafts being revieiwed, it is considered important that a thrid part be asked to fill the shepherd role. After the meeting, Dow Street agreed to take on this role. He will be working with the RRG and the IRSG to handle an issues related to the drafts. ================================================ Brief notes on comclusions and discussions after presentations ================================================ Elwyn Davies Brief Update on the IAB Routing and Addressing Workshop - http://www3.ietf.org/proceedings/06nov/slides/RRG-0.pdf A review of the IAB routing workshop which covered not only the problems but the areas in which the the research community can contribute. Appropos of this topic, a decision was made at the end of the meeting to schedule a full day RRG session on this topic for the Prague IETf meeting. While details remain to be worked out, the attempt will be to schedule this meeting for the Saturday before the Prague IETF meeting - 17 March. Among the issue that Elwyn brought up for discussion are: Discussion: what can research community do to help? Elwyn's personal thoughts: - there has been too much fixing of symptoms -- need to determine sustainable future rather than point fixes - very difficult -- ossified thinking, least common denominator thinking - what really is the scaling problem? -- power issue is serious -- not just a pure routing problem -- interaction w/ packet classification and ACLs is important - routing network topology is changing - BGP tools not so effective as they were - need to support TE in more integrated fashion -- controllable and manageable mechanisms - meaning of identity -- makes sure disentangling of semantics solves the right problem -- usually we overload the current name space w/new semantics rather than other approaches, because latter cost money -- have to think about treating these costs as an investment in business growth - implication of separating identity and location: need second look-up -- overload DNS? - is there a cheap, fast, non-hierarchical, scaleable, distributed way? -- current view that we have exhausted possibilities, but historically new possibilities have appeared - is there failure of imagination? Question and comment: what would be the scope of identifier -- affects where and when extra lookup happens. Short term fixes: - improve iBGP - major problems now - tools to help AS to apply policy from centralized point -- reduce instability, reduce demand for skilled BGP hackers The presentation included pointers to further discussion. Avri: please continue discussion on RRG list. Margaret: excellent summary. Who is paying? She thinks enterprises are. Can't get PI addressing or are paying a lot for it. IPv6 a red herring. Does wonder: why does same hdw have to route IPv4, IPv6? Was this discussed in workshop? Very little. Answer: Comment: need to evalaute tools against rock-solid models -- need to focus research on latter. Comment (Lixia): ?? ======================================================= Routing in Next generation http://www3.ietf.org/proceedings/06nov/slides/RRG-2.pdf Jordi Palet RiNG project - surveying ISP and user reqts for routing in NG ntwks Variety of European participants Looking to provide input to IETF, IRTF, other fora References: http://www.ist-ring.eu http://wiki.ist-ring.eu Avri: lots of European students looking for projects -- hope RiNG can help them ==================================================== Aaron Jaggard Formal models for Path-vector protocols http://www3.ietf.org/proceedings/06nov/slides/RRG-4.pdf (bulk of these notes may be remived once the slides are submitted) Presenter looking for: - areas of collaboration - feedback on previous/ current work - questions for future work Questions how to refine work to make it more useful? - class-based routing an usable model? - should focus be on next-hop w/tweaks? - what are pressing theoretical questions? Are there related areas for joint work in RRG? What sort of policies should they be looking at? - what are feasible constraints? - what sort of policies are usually written? - are there shortcomings in expressiveness? - what sort of reconfiguration capability is needed? - What sort of policies are typically written? - What anomalies do people see? Are they seen often? - What are typical iBGP policies? Avri: asked that these be propagate to the list for further discussion and recommended that other questions be sent to the list as they come up. ==================================================== Pierre Francois Toward Harmless Maintenance Operations in IP Networks http://www3.ietf.org/proceedings/06nov/slides/RRG-1.pdf while the iBGP case is being worked on in the IETF RTGWG and is more engineering then research, there may be research work in the eBGP case. ==================================== Scalability of routing, compactness and dynamics http://www3.ietf.org/proceedings/06nov/slides/RRG-3.pdf Daniel Krioukov Conclusions: - routing table sizes can be made extremely succinct - but all known algorithms require all nodes to have full view of the network topology graph - and info on changes has to be propagated promptly We can help by - increasing awareness of problem -- need knowledge that is spread about in many places Rudi: as operator, gets from this that limited diversion needed from shortest path. In practice, have to respect constraints on path capacity, economic and social constraints. Latter: operators have to feel they can control the routing in the way they want. Response: so full view isn't enough. ==================================================== Discussion: so are people interested in this work? Should RRG continue? How many have read reqts draft, how many think it should be published? (About a half dozen in each case) Is the IRTF an useful place to do research work in this area? Respondent feels increase of interest. Avri: could additional mtgs be scheduled, more than once a year? Focus on work done on questions Elwn brought up. Are there people interested in bringing work to next IETF? Saw 8+ hands. Aaron Falk: trying to get groups to have at least one full-day mtg a year. How many would come to such a meeting? (About a dozen). Rather more if tacked on to IETF meeting (before or after). Does the continent matter? (No) Avri will take to the list for follow-up.