IETF 91: Open Shortest Path First (OSPF) WG Agenda Wednesday, November 12th, 2014. 13:00 - 15:00 Afternoon Session I Location: Lelua Suite ======================================================== Chairs: Acee Lindem Abhay Roy WG Status Web Page: http://tools.ietf.org/wg/ospf/ 1) Administrivia - 5 minutes - Blue sheets - Scribe (Les Ginsberg) - jabber (Dhruv Dhody) - Jabber room: ospf@jabber.ietf.org 2) WG Status Update - 10 minutes See Slides Anton:draft-smirnov-ospf-xaf-te OSPFv3 instance can use a tunnel whose attributes are signaled by another OSPF instance or AF. Better handles corner cases than other solutions. Please send comments. 3) OSPF Router-Information LSA - RFC 4970Bis - 5 Minutes - https://datatracker.ietf.org/doc/draft-acee-ospf-rfc4970bis/ - Acee Lindem See Slides Shraddha Hedge: New draft has functional and info capabilities. For functional capabilities would it make sense to make this MT capable? Acee: Please send email to the list on this. Shraddha: Should we include MT discussion on all new work? Acee: Take it to the list 4) OSPFv2 Prefix/Link Attributes - 10 Minutes - https://datatracker.ietf.org/doc/draft-ietf-ospf-prefix-link-attr - Peter Psenak See Slides Acee: Since there are multiple implementations should be able to do WG LC before Dallas. Andrew Dolganow: Agree should go forward since there are multiple implementations. 5) OSPF/OSPFv3 Segment Routing - 10 Minutes - https://datatracker.ietf.org/doc/draft-ietf-ospf-segment-routing-extensions - https://datatracker.ietf.org/doc/draft-psenak-ospf-segment-routing-ospfv3-extension - Peter Psenak See slides Acee: Happier with separate advertisement for ranges since ranges will have a subset of attributes compared to prefixes. 6) OSPF Yang Data Model - 15 Minutes - https://datatracker.ietf.org/doc/draft-yeung-netmod-ospf - Derek Yeung See slides Acee: Chatty notifications (e.g. stale LSA version due to restart) omitted from MIBs - also omitted here - should they be? Jeff Haas: Notifications - use knobs to enable/disable Acee: Similar proprietary extension exists in OSPFv2 MIB implemenatations. Jeff: Instances are abstract - don't know what they map to. Derek: Instances map to address-family Jeff: Being worked on in NETMOD. Derek: Current model allows it to adapt. Jeff: Topology model in I2RS undecided. See if these models can be normalized. Acee: Will that be presented in I2RS? Jeff: Not this time - will be pickup shortly Dan: Authentication - could be part of security - not in OSPF model Acee: We have folks working on separate keychain mode Derek: OSPF model will refer to that Wenhu: Address family under interface level. Since model support V2/V3 should AF be kept under instance? Derek: Thought we removed AF from interface level. Acee: Dual stack exists in the model - though this is not yet supported by the OSPF protocol. Will align with IS-IS which does support it Derek: It is a feature - not yet published 7) I2RS OSPF Yang Data Model Additions - 10 Minutes - https://datatracker.ietf.org/doc/draft-wang-i2rs-ospf-dm - Sue Hares See slides Jeff: Please think about whether there are things that are not part of "standard configs" which you might want to do Sue: See use case draft in I2RS Acee: How many have read it? (About half) Many things are R/W that are unexpected (e.g. Local RIB). Sue: Goal in I2RS was to follow the use cases Want input if all use cases are good ideas. Not comfortable with modifying OSPF DB (for example) Please provide feedback Dan Bogdanovic: Started by exposing everything. Met resistance. We don't know what is safe to expose via I2RS interface. This is true across all objects (IGPs, firewalls) etc. Can create instability. Rob Shakir: I2RS use case draft is a big basket. Is it sane to do it? Why would operators want to do some of these things? Sue: Some BGP operator input being received in I2RS - would like to get input from OSPF WG. Want more input on what makes sense Derek: For example, local RIB is missing in OSPF will probably be added later. Sue: Is feedback on Local modification a bad thing. Derek: Should be read only. But differences with I2RS may diminish over time Are you asking more R/W support in OSPF model? Sue: No - tell us if I2RS use cases of R/W make sense in all current cases. See ephemeral state discussion. Hannes Gredler: having programmatics APIs .... so computers programming routers and thats plugged straight into a flooding protocol machinery - well, that fits my definition of a "weapon of mass destruction". please *do* remove write access to the LSDB - it's a really bad, bad idea. Dan: Please get right definition of ephemeral state. Empemeral state vs emphemeral config. At Juniper cannot support emphemeral config - only ephemeral state . Jeff: Goal for use case is historical data point - please provide feedback on what is sane. Won't go forward w/o incorporating feedback. Acee: Who has read YANG model draft? (More than use case) Point of presentations is to try to reconcile the definitions and poll for WG adoption. 8) OSPF Topology Transparent Zone (TTZ) Prototype - 15 Minutes - https://datatracker.ietf.org/doc/draft-chen-ospf-ttz - Huaimo Chen See slides Tony Przygienda: You are building backbone areas w TTZ. Area 0 is unique which preserves loop free properties. You are building ASs - which means you will have looping. Acee: Looks like a mesh of border routers - internal routers are hidden Tony: Need to rethink Hannes: Reducing network paths is not always desirable. (LFA for example) You are reducing FRR coverage. For TE (e.g. SRLG admin groups, RSVP resolution schemes) becomes more difficult. Not discussed in TTZ draft. Huaimo: Have TE TTZ draft ready (not yet submitted) Acee: Protocol extension is possible. Yet to agree when it is useful. Poll for adoption as Experimental Tony: Problem is valid (hitless Area/TTZ migration). Should be worked on. 9) OSPF Stub Neighbor - 15 Minutes - https://datatracker.ietf.org/doc/draft-raza-ospf-stub-neighbor - Faraz Shamin Jon Mitchell: Poor choice of draft name. Draft is a hard read - not clear what the router mechanics should be. Faraz: Hub is originating 2 router LSAs. Local one should be be ignored. Jon: How does other hub know it should ignore "local one? Faraz: Will clarify in future draft Jon: Similar to EIGRP Stub. What should hub do w external - pass through? Faraz: Will think more about this. Shraddha Hedge: Typically have layers of hub and spoke - how will you address this? Faraz: Only talking about lowest level Shraddha: But hub can be spoke for upstream Acee: Hierarchical topology is out of scope. David Lamparter: Looks a different routing instance at the hub. Could do the same with multiple instance Acee: People don't want to manage that many instances. Heavyweight solution. Requires policies between instances. Could make each sit eits own stub area - but this could require thousands of areas. Jon: Protocol mechanics have to be defined. Current defiunition of draft is too restrictive. Wenhu Lu: Why does this need to be standardized? Faraz: Sending default route which is non-standard (differs from RFC 2328) Wenhu: Is there change for stub routers Faraz: RFC 2328 does not reserve 0/0 for stub link advertisement but Cisco doesn't allow it so some stubs would need a change to allow this. Acee: Early in this work. Agree protocol mechanics not well defined. 10) OSPF Extensions for Flow Specification - 10 Minutes - https://datatracker.ietf.org/doc/draft-liang-ospf-flowspec-extensions - Jianjie You See slides Presented by Eric Acee: There was another draft that proposed the same thing - use cases did not justify going forward. BGP supports this - with IGP the flow Spec goes everywhere but not needed everywhere. Do operators believe this is useful? Discuss on list 11) OSPF Node-Level Admin Tag - 10 Minutes - https://datatracker.ietf.org/doc/draft-hegde-ospf-node-admin-tag - Shraddha Hedge See slides Acee: Respin of 4964 will make 64 tag limit a non-issue. Acee: New version reflects comments received during WG adoption. 12) OSPF Self-Defined TLVs - 10 Minutes - https://datatracker.ietf.org/doc/draft-chunduri-ospf-self-defined-sub-tlvs - Uma Chunduri See slides Hannes: Name is confusing - like IS-IS GENINFO Uma: GENINFO must be non-zero Les Ginsber: Application ID is non-zero. sub-TLVs are application specific and documented/assigned elsewhere Hannes: If multiple vendors implement this then standardization process is helpful Uma: Yes - we deliberately want to bypass standardization Hannes: experimental? Uma: Experimental is a possibility Alia Atlas: Is the purpose to bypass request to IANA? Uma: Codepoints may not be IANA. Alia: All values could be experimental or first come first served. Uma: Think of this as extension of admin tag. Hannes: Then use admin tag. Alia: Experimental? Rob: Is it opaque to OSPF? Assuming you need interpretation could be disastrous. Assigning codepoints is not the blocking issue Acee: Need to discuss on list on why we need this?