RTG area IETF-59 Intro Bill: Routing Area Web page fallen into disuse New routing area directorate folks to replace Tony and Tony. New routing working group as catch-all Close BGMP, IDMR - Saying this for 2 years - Now date is set to wind them up (July 2004) - Real target! Create BFD WG before San Diego Create RPSEC WG especially for BGP security around October. - Requirements are coming somewhat from other SDOs. Dave: What is the disposition of the BGP draft? Bill: See slide later WG reports BGMP - after approval as experimental found bug. - no AS number in BGP open message, but text said to use it! - new revision planned Dave: Will we have no stds track IDR protocol for multicast? Bill: Yes, that's what it means Dave: SSM makes stds track then this means we endorse SSM as *the* multicast IDR for IPv4 Bill: Good point. But since scaling problems with one, and almost no interest in BGMP Dave: Different to say no interest in BGMP and no interest in multicast IDR Bill: But clearly no interest in scalable protocol Maybe this means there is no interest? Dave: But there is interest. And experimental protocol is deployed. Have been asking this for some time. This seems to be a statement to solve implicitly using SSM Bill: Don't think this is what we're saying. Peter: There is ASN traffic in the n/w MSTP does the job today. Has problems but not seen them. It is holding together and being used. Try to figure out the growth of MSTP in the Internet and try to predict when we will hit the problem. Someone needs to figure a protocol to address all required functions Bill: Hard to say where effort to figure out IDR multicast should go WG charter doesn't help. cf IDMR WG. Takes more than a place to do the work. Takes energy and ideas and feeling is that those aren't coming. CCAMP Adrian: No questions. IDMR Bill: Dave Thaler: what is the status of DVMRP. no security Bill: pushing for proposed std track since it has uses in certain environments IDR Yakhov: Finished base spec. Very significant. Many docs now go to IESG Can now progress other items Can now start new work - cannot pick it all up - need feedback for priority list - SPs and vendors - please input Sue: Thanks for BGP implementation reports respondents Thanks to the people who processed the 260 questionnaires ISIS Alex: New co chairs for ISIS Should ISIS stds go to stds track? IGP pt-2-pt pending revision OSPF Alex: graceful restart recently approve v3 authentication and MIB update to ADs 2547 extension for loop prevention minor comments from IESG interesting discussions on mailing list multi-address families, and multi-area adjacencies MANET Alex: DSR spec is pending update after AD review RPSEC Alex: Needs more input on requirements. Please do help. Security requirements for BGP security (see previous new WG) Routing WG Alex: New WG No documents yet. Wrong! MPLS/SPF for traffic engineering - expect comments and last call soon Please subscribe to the mailing list rtgwg-request@ietf.org MPLS Loa: Had meeting Monday Most cycles on TE - P2MP - OAM framework A couple of docs about to IESG or with updates after review (7 or 8 of these) 12 docs in RFC ed queue about to be released as references are sorted out ??Topic?? Bill: Spec updated to fix a blackhole Means spec was reviewed very well! IESG in a month or so SSM Some delay while discussing Apple(?)'s claim on IPR Believed to be resolved VRRP Bill: v2 approved v3 (IPv6) - some issues found - most resolved Routing Research charter Avri: see slides Closed groups are going to do quarterly reports to the list and an annual gathering Research proposals please - short term (for example as rejected by WG) - complete the work in 6-18 months - longer term - e.g. is there life after BGP? - e.g. routing and protocol scalability This is research so if the end result was "bad idea" this would still be good research. meeting slot planned for ietf 60 interim (co-sponsored) meeting is planned Alex: How to get onto closed mailing list? Avri Send to main list asking for "in" Alex: How do you get deliverables from a research group Avri We have to work with them to set up milestones Keep open line between rtg area and rtg working group IANA Alex: Who read the draft? answer ----? see slides Issue - deployment of drafts uses interim codepoints Often considerable deployments Other non-rtg similar issues. AFI/SAFI Sue: see slides Used *before* reach problem stated by Alex Kireeti wants an AFI Alex: Thanks to Kireeti Bill: RFC3692 on experimental alloc suggests pick a small set and designate them for experimental alloc default behavior is non-support - needs to be turned on by operator Bob: IPv6 had discussion on ICMP codes ended up with 3 options could use extendible codes IPv6 leant towards alloc when adopted by WG - becomes permanent when RFC Need a generic way not just rtg-area Alex: that is intention Bill have you adopted any ideas? Bob: No, beginning to put language in Bill: in case of temp alloc, concern is you assign to a proposal, but WG now needs to massage packet format, but it is already deployed with a number, what do you do? new number? Bob: Up to WG chairs as per Alex What about individual submissions? Alex: This doc tries to address stds action policy not individual - would need to go through IESG Kireeti No early allocation for individual drafts Bob: Same problem for ICMP etc because this is "IESG action" Kireeti: Alex and Kireeti is stds action for early alloc Sue is talking about experimental which might go to stds track Alex: Would this be useful to IPv6 Bob: Yes WG would consider using it twiddling values in experimental by end user might be not good Kireeti: This is experimental only Bill: Would not be good for major deployment Dave: In last proposal not clear whether to assign a range but not a specific number, or if WG says specific number. Bill: RFC doesn't say Dave: Not asking about RFC. What about configurable for experimental. Is usage agree ahead of time, or is it up to the implementer to force operator to type it in? Bill: Either. must be config because it is from experimental range and someone may have conflicts Dave: Compare different proposals. like subrange == experimental Or mark as temporary out of ordinary == good because no change in implementation if no changes Bill: disadvantage if you rev or peter out. cannot reclaim until timeout Dave: Bob and Brian, 27 allocated over ten years Time limit may not be important in large space Kireeti: Not all codespaces (eg IP protocols is implicit stds action) have explicit actions associated Proposal is specifically for stds action space Need stable allocation Experimental is not a big deal if there is a change First come first served is dangerous in a different way Dave said what does one year mean when it expires? need to make this clearer. Early is in some sense temporary If spec dies it becomes "deprecated" and then time that out (e.g. RSVP case, value was not marked deprecated so IANA re-allocated) Alex: Sense of the room, about draft-kompella-zinin approach is useful? "pretty good" approach is not right? - just Kireeti ;) ditto Afi/safi useful? very few not useful? none need to ask in IDR IP/LDP loop protection Alia: see slides pre-compute an alternate path cf. IGP fast reroute questions after second presentation next hop fast re-route Naiming: see slides Yakhov: slide 3 R1 R2 is pt to pt, but suppose multiaccess media? Alia: if all nodes in same area, and R2 is the primary neighbor then R1 can assume traffic is from R2 Yakhov: If SPF takes zero time then would there be any value Alia: propagate is non zero install time is non zero Yakhov: not need to propagate to all Alia: Some cases there may be more than one router Yakhov: If only one? Alia: If more than one then it is valid Kireeti: 80% if no U-turns? Alia: 79.5 Kireeti: Real network? Alia: Yes Kireeti: U-turns in real networks? Alia: Yes Cengiz: I tried without u-turns even then in the networks I tried I had 90% Kireeti: I'm surprised that the number is so low w/o uturns because network is dense Alia: topology dependent analysis based on src/dest pairs lots of traffic not covered by loop free percentage of traffic is not same as percentage pairs Rahul: what if equal cost paths to neighbor Alia It works Rahul: R2 is PLR R1 u-turn neighbor How does R1 know where to put traffic Alia: If there is only one primary neighbor [note taker got confused ....] Rahul: when I said doesn't work I meant reduces Alia: never reduces below loop free w/out utirn Yakhov: Covers node failure? Alia: Yes Ina???? Deployment considerations in n/w Alia: Need to support IGP extensions for u-turn ??? How fast Alia: depends on h/w per interface forwarding table Kireeti: What effect on the size of the routing table A Forwarding table not routing table Kireeti: bad? Alia: Not as bad as one per i/f since max two Kireeti: has naiming done coverage? Alex: take discussion to RTGWG (not routing discussion list) RIFT David Meyer: slides send questions to the GROW meeting OIF routing Lyndon Ong: there is liaison statement - need to find out relationship to get people access to it OIF has been working on E-NNI routing specification based on OSPF model for transport n/w is similar to ITU work elements grouped into domains and connected to each other through E-NNI reference points some domains may use distributed control and others may use EMS but still interface to other domains through sig + routing protocols hierarchy domains == RA (routing area) model is similar to ITU model domains may be partitioned to give mulit-level hierarchical relationship Assumptions see slide Status see slide testing at OFC in 2003 did not test any hierarchy extensions not a std solution defined as prototype for testing Rahul: concerning to see changes to OSPF without telling OSPF WG hierarchy is alarming I'm expressing concern Lyndon: recognise concern hierarchy is not simple and not something we have been testing bringing it here now Dimitri: How does this differ from CCAMP DT routing for ITU since that work follows ITU model Why take this experimental solution and not develop based on DT Lyndon: problems here are consistent w/ problems in DT difference here is that OIF has gone ahead and defined what some of these extensions might look like Russ: please give ptr to url Alex: liaison will be available Kireeti: CCAMP chair hat on worked on it for a while and bring it now say this is for testing which is good but now you have brought it to us if ask for rubber stamp won't happen but you need codepoints and we want a good relationship tell us the requirements, and we will extend since we know protocols big risk of two sets of solutions that are different compare with RSVP work where OIF/ITU work is not compatible protocols are understood here where implementers and deployers are found are you asking for rubber stamp what is end goal? Lyndon: not approved through OIF as an implementation agreement it is a draft if we can work out how to do it in OSPF then it is not too late Kireeti: do you want us to take your changes do what with them? Lyndon: oif did send a liaison asking for comment/feedback sent to ccamp chairs and routing area directors expectation is not rubber stamp expectation is that this group have a look at the problems that generated these prototype extensions hope is that these problems will be addressed through std extensions of OSPF whether the extensions change is up to this group Kireeti: would have been better to bring requirements Arthi: we can work through the problems and have solutions to them and get discussions Lyndon: work wasn't done for the sake of it people looked at specs and said "we don't know how to do something" Zafar Ali: need to understand overlap with ITU requirements follow same model Lyndon: should look for delta to ASON requirements Zafar: if so we can overlap the solution space Dimitri: how much is change to basic protocol and how much to TE LSAs? how has OIF evaluated impact analysis Lyndon: expertise for impact is in this group changes were to achieve specific functions Alex: there is concern about OSPF extensions being defined elsewhere your answer is your not trying to define a std your intention is to experiment and see what the extensions might look like your intention is to come here and work with us apology to Christophe - out of |