Routing Area Open Meeting (rtgarea) IETF 94 (Yokohama) =============================================================================== Area Directors: Alia Atlas (akatlas@juniper.net) Deborah Brungard (db3546@att.com) Alvaro Retana (aretana@cisco.com) Area Secretary: Jonathan Hardwick (jonathan.hardwick@metaswitch.com) Wiki: https://trac.tools.ietf.org/area/rtg/trac/wiki/WikiStart Scribe: Jonathan Hardwick (jonathan.hardwick@metaswitch.com) Location: Room 501, Pacifico Hotel, Yokohama, JP Time: November 2, 2015, 1710-1910 (5:10pm-7:10pm) ------------------------------------------------------------------------------- Alvaro opened the meeting at 17:15. ============= 1. Administrivia ---------------- As well as a YANG technical report there will be a technology presentation on RPL. 1 new WG - detnet. No WGs closed or rechartered. New chairs: detnet {+Lou Berger, +Pat Thaler}, pce {+Jonathan Hardwick}, bfd {s/Nobo Akiya/Reshad Rahman/} 2. Routing YANG Technical Coordination Report --------------------------------------------- Qin Wu. Base models should provide a common structure for all other models. Model extension can extend in two ways. 1) Augment from base model re-uses common structures and adds technology specific additions. 2) Reuse common building blocks from the model. More information on IETF YANG can be found on the YANG wiki. Use it. There are 20 IETF YANG drafts in the routing area. Many more individual submissions. What is missing? (Bill's opinion only.) - Base model for tunnels - Base model for QoS There are some issues where YANG models are currently standalone but should augment a base model. Please keep the wiki up to date. It would be useful to have an automation tool to generate the relationships between modules. Dave Sinicrope: Note also, from PALS meeting, Benoit Claise create a catalog of YANG modules in other SDOs. We should link this to the wiki. RTG Technology Presentation "RPL Overview" ------------------------------------------ Michael Richardson / Ines Robles Here is a summary of some key points about RPL. Please see the uploaded slides for details. RPL is pronounced Ripple! RPL is used by constrained-resource devices (with similar compute power to a CBM64) that operates on a low power source over a lossy network. RPL is a distance vector, source routing protocol. RPL organizes the network as a Directed Acyclic Graph (DAG). The network is partitioned into one or more Destination Oriented DAGs (DODAGs). DODAGs are "grounded" or "floating". If a node loses its connection towards the root of a DODAG then is detaches itself from the grounded DODAG and becomes a floating DODAG. Nodes that are detached try to reattach to a grounded DODAG. Each node maintains its rank in the DAG (distance from root) and must rejoin a DODAG to a node with a lower rank (which prevents loops / nodes joining to their children). The protocol messages are ICMPv6 messages. RPL supports MP2P, P2MP or P2P flows. MP2P is the most common in applications (e.g. collect data from a sensor network). RPL nodes support virtual routing via multiple instances of the RPL protocol. RPL has different metrics - and these are extensible. For example - Node state ~= current workload. - Node energy ~= how much power has a node got, could be expressed in terms of e.g. how many more packets can I forward before my battery dies. - Link throughput. - Link Latency. - Link Readability. - Link Admin Cost. The metrics go into objective functions. Only two are currently defined. Again this is extensible. The plan is for researchers to work out what are the best metrics and OFs. ROLL did an extensive survey of existing routing protocols before deciding to create RPL. There are several implementations of RPL including some that are open source. There are variants of RPL that build in mobility of the devices, which has applications to IoT. Jeff Haas: What are the security properties of this protocol. Michael: I asked that question and ended up as co-chair! Anyone who is deploying is doing so on top of a L2 that they think is secure. Strategies for adding nodes at L2 are proprietary. Not interoperable. There is a consortium outside the IETF that is creating a protocol that allows you to join at L2. The IETF is defining packets that allow asymmetric key signing of update messages but this is beyond the capabilities of existing hardware. Charlie: When you evaluated ODV, one of the objections is it did not support upcount, plus . Michael: I was not involved at the time, but the conclusions are in the draft. Charlie: Is it true most implementations use non-storing mode? Michael: Contiki supports storing only, but deployed networks (which tend to be closed and proprietary) do not support it. Charlie: What is the traffic density in the 5000 node networks? Michael: Collect a datagram per metre per hour. Data rates are packets per minute. Charlie: Disadvantage of non-storing mode, if you source route, its more expensive to transport the additional data than to do a bit of extra computation up front. Michael: Broadly agreed. Alvaro cut the discussion. Working Group and BoF Reports ----------------------------- Routing Directorate Report WG Chair Reports Alvaro: please make sure you have some text in the wiki to highlight what is going on in your WG. You must do this before the RTGAREA meeting even if you have not met yet. You should update the wiki at the end of the week when you have met. BESS WG (chairs cannot be here) BFD WG (Jeff Haas) Nobo is too busy so has handed over to Reshad as chair. This week we will cover BFD for VxLAN. Also BFD YANG and fast authentication - does more sustainable crypto, as it's impractical to sign packets on many sessions every 3.3ms. One RFC about to be published, one draft through WGLC, two currently in WGLC including 1 MIB. BIER (Alia) Making good progress, busy, meeting Friday CCAMP (Daniele Ceccarelli) Four drafts in IESG queue. Black links work has stabilized. Dealing with incoming liaisons from BBF and from ITU-T. DETNET (Pat Thaler) First meeting of the WG. We have a consolidated use case document that should soon be adopted by the WG. Two other documents we are working on in order to adopt: problem statement and architecture. We have started discussions on the content of the data plane. I2RS (Sue Hares) 8 documents are ready to go to the IESG. They are finalizing the initial data modules. Issues: Should the RIB model be aligned with the ROUTING-CONFIG model? FB-RIB is a feature for config and ephemeral - but there is no config. Initial protocol proposal is RESTCONF. IDR (Sue Hares) 3 new RFCs, 2 drafts with RFC Ed, 3 drafts with IESG 7 drafts that will either be pushed over the finish line or abandoned. The focus for this meeting will be flow-spec. IS-IS (Chris Hopps) Discussing: - Change to Router Capability TLV to remove dependency on IPv4. - Change to LSP Remaining lifetime. - Advertising L3 Bundle Attributes. 4 drafts have passed last call. 2 drafts have been submitted to IESG. L2TPEXT (Carlos) Not meeting this time. Acting to clear the queue of documents. LISP (No show, not meeting) MANET (No show, not meeting) MPLS (Loa Andersson) Only one chair in Yokohama, so Loa is flying solo. Meeting twice this week, including a joint meeting with TEAS to discuss YANG. LSP-PING and LDP are ready to become full Internet Standards. The YANG work is ready for WG adoption. NVO3 (Benson) Still 3 data plane drafts. A DT will be put together for YANG model. A DT is up and running on OAM. Discussing a physical interim meeting in Feb timeframe. One draft has WGLC but requires further review. Two drafts ready to enter WGLC. OSPF (Acee) Prefix and Link Attributes work is finished. Segment routing nearing WGLC. YANG module is feature complete. PALS (Andy Malis) Have already met. Talked about YANG for L2VPNs. 3 drafts with RFC Ed. 5 drafts with IESG. One is intended for Full Internet Standard. 2 drafts will go to WGLC after the IETF. PCE (Julien) Welcome Jon Hardwick as a co-chair. 2 drafts have gone for early code point allocation. 1 draft sent to IESG. 3 IDs getting ready for IESG after one new revision from authors. 1 draft passed WGLC. Several drafts to do with stateful PCE are being discussed. PIM (Stig Venaas) 1 draft in RFC Ed queue. 2 drafts with IESG. Several new proposals have come in, including a YANG proposal. There is a YANG DT that is making good progress. ROLL (Michael Richardson) 3 documents in RFC editor queue. 2 drafts ready to be submitted to IESG. 5 new drafts, 2 will be discussed this week. RTGWG (Jeff) Have met once, meeting again. Good progress on YANG drafts. 1 new RFC. 1 draft with RFC Ed. SFC (Neither chair is here, Carlos and Joel will chair it) Architecture is RFC Control plane document is adopted SIDR (Sandra Murphy) Meeting twice. 1 doc in RFC Ed. queue. 1 doc in IETF last call. 3 docs passed WGLC. 2 docs that were revised after WGLC and will be sent to IESG. 1 new document now adopted. BGPSEC went through WGLC but a late breaking comment will need a change and then a new short WGLC will be issued. Validation Reconsideration is on the agenda but some question on whether the author will continue. SPRING (Bruno) To discuss: - Anycast segments - Error handling esp. segment ID conflicts - Handling multiple algorithms: should it be multiple SIDs or multiple SRGBs? 1 draft submitted to IESG 1 draft has passed WGLC 4 drafts adopted TEAS (Vishnu) Two sessions this week including a joint session with MPLS to discuss YANG. To discuss: - TE topology YANG model. - RSVP/TE YANG models. - ACTN. 2 docs in RFC Ed queue 2 docs with IESG 4 docs are adopted. TRILL (Sue Hares) Congratulations to Jon Hudson (co-chair) who is not here because he has a new baby at home! 5 drafts in RFC Ed queue 5 drafts passed WGLC 13 drafts are in the WG. Open Discussion / Any other business ------------------------------------ Use case documents. Alvaro recently recommended a WG not publish a use case document. It was just a personal opinion not a AD directive about use case drafts. Why? There may not be value especially if it is coming after the protocol specification. New use cases for existing technologies turn up all the time but there is no need to publish them. The original use cases may date badly. NB there is no general policy, but Alvaro will continue to express his opinion. Alia: we should focus on describing how to implement and deploy. Use cases have a limited shelf life. If the ADs push back it is to discourage the WG from spending non-useful times. Jeff: We agree. Use cases do not generally make useful RFCs. But it is often useful to write down the use cases when you start on a piece of work, and someone has to put in the effort to do that. Given we had got so far with this document, and had invested lots of effort. Why not publish given where we had got to? People gave their time freely. If it is the IESG's general policy not to publish use cases then let's say that to prevent further time wastage. Secondly, let's provide some recognition to people who worked on use cases, even if we don't publish. Alvaro: for now there is no change to the official position, but we will push back to say if you are going to publish this, at least make the quality sufficient for an RFC - and consider whether it is worth the effort to do so. Alia: if use cases are published, it should be a short and non-exhaustive summary. It should not be done simply as a reward mechanism. ======== Alia closed the meeting at 19:07.