Babel WG - IETF 108 VIRTUAL - Monday, 27 July 2020 13:00-13:50 (UTC) Room 2 Chairs: Russ White (Juniper) Donald Eastlake (Futurewei) Notes: Gabriel Kerneis (edited by Donald Eastlake) Area Director: Martin Vigoureux (Nokia) Agenda ====== (times are as scheduled, not actual) 3 min. Administrativia (scribes), Agenda Bashing, Chairs 5 min. Status, Milestones, Chairs 15 min. Babel IPv4 over IPv6, Théophile Bastian draft-bastian-babel-v4ov6-01 10 min. Babel Information Model, Barbara Stark draft-ietf-babel-information-model-10 5 min. Babel YANG Model, Mahesh Jethanandani draft-ietf-babel-yang-model-06 9 min. Babel Milestones & Charter, Donald Eastlake 3 min. Wrap-Up, Chairs Status, Milestones ================== Donald Eastlake (Futurewei, Co-Chair): [See slides: https://datatracker.ietf.org/meeting/108/materials/slides-108-babel-agenda-and-status] Reviewed agenda and mentioned that the first presenter was having problems connecting so we may move their presentation later in the agenda if there is no objection [there was no objection]. Working group document status reviewed. Almost all milestones completed; upcoming proposal (later today) to add some more. Juliusz Chroboczek (IRIF - Univ. of Paris-Diderot, jch): IESG changing, even further delays? A single AD has held up the Babel draft for a year now, what is the impact on the document of the IESG change? Should we give up? Donald: There was a turn over in some IESG positions in March. We know how to clear the objections from this AD [Area Director]. We don’t give up yet. jch: I am worried about the amount of energy required to discuss with the IESG. My contract to support Babel has expired. David Schinazi has changed companies in the meantime. David Schinazi (Google): There have been unfortunate delays, but we have a path forward. Action item: make the latest edits, submit the changes, that would allow Alvaro to lift his discuss and then expect it swift approval from other ADs. jch: How do we make sure the process terminates? Do we decide on a deadline? David: We are chartered to create a standard and need to complete the process, to get a standard track document. jch: Tired that it takes so much time. David: Agree that IESG has taken some time, but there is also some work needed from the authors. Not sure how to set a timeout, let’s fix the document first. Donald: The next item on the agenda is Babel IPv4 over IPv6 but Théophile is still having problems connecting. So, if it is OK with you Barbara, can you go to the next item on the Babel Information Model. Babel Information Model ======================= draft-ietf-babel-information-model-06 Barbara Stark (AT&T) presenting. [see slides: https://datatracker.ietf.org/meeting/108/materials/slides-108-babel-babel-information-model-draft-ietf-babel-information-model] 2: Mostly editorial edits recently. A few items to clear. 3: clearing vs. enabling statistics. 4: clearing vs. enabling packet log. Trickier than the previous one, would like input from implementors. Unclear what sort of behavior we want to get. 5: Mahesh argues that YANG cannot do much of anything, but sounds odd to Barbara. 6: What are the thoughts of other implementors about packet logs and clearing them? [nobody chimes in] Not sure what to do in the absence of input. Need to keep it very simple if nobody is implementing this right now. 7: converge-fast Chattiness vs. convergence speed tradeoff setting. Do we want a boolean to pick between the various strategies? jch: The proposed converge-fast is essentially a macro for a set of parameters. Expresses a user choice, but it is redundant information for the other parameters. Eg: what if converge-fast=true && hello-interval=1 hour (which is contradicting). barbara: Those other parameters (eg. hello-interval) are read-only as was discussed. jch: That is unexpected to me. barbara: It was complex to allow users to change them, that’s why we decided that. jch: Let’s take this offline. 8: Will we need another WGLC? Donald: I think these adjustments are minor enough that if they are discusses on the mailing list another WG Last Call is not necessary. Although this document is in "publication requested" state, I believe we can still update it. Donald: Since Mahesh has a conflict, can you do his brief YANG presentation Barbara? Barbara: OK, since we are here in the agenda anyway. Babel YANG Model ================ draft-ietf-babel-yang-model-06 Barbara presenting (Mahesh had a conflict with this meeting). [See slides: https://datatracker.ietf.org/meeting/108/materials/slides-108-babel-babel-yang-model] 1: Have been and will continue to update YANG model to match changes to Information Model, no problem. 2: Babara happy with the choice of not doing anything for item #15, simpler solution. jch agrees on jabber. 3: Don’t expect to have any more comment. Read for WG Last Call? Donald: There was a WG Last Call on this document that was never closed out and is stale at this point. We will re-start WG Last Call on the YANG document. Babel IPv4 over IPv6 ==================== draft-bastian-babel-v4ov6-01 Juliusz presenting (Théopile still having technical problems connecting). [See slides: https://datatracker.ietf.org/meeting/108/materials/slides-108-babel-announcing-ipv4-routes-with-an-ipv6-next-hop-in-the-babel-routing-protocol] Work by Théo, funded by Nexifi, supervised by jch. 1: hidden assumption of a routing table: IP4 address -> IP4 next hop, IP6 address -> IP6 next hop, but never IP4 address -> IP6 next hop. 3: In IP4, next hop address is only used to get a MAC address of the next hop. It never appears on the wire. So why can’t we have a IP6 next hop, even for a IP4 address? 4: It’s possible, introducing the concept of a v4-over-v6 route. (not very happy with terminology, because it evokes tunneling whereas it doesn’t use any; looking for suggestions of better name) Same idea used in a draft=ietf-bess-rfc5549revision for BGP. 5: Software routing -> Linux supports it since July 2019. Hardware routing -> don’t know if there is any vendor supporting it. 6: Description of implementation in babeld by Théo. Changes required only for IP6-only interfaces. [At some point in here Théo succeeded in connecting but asked Juliusz to continue with presentation] 7: With this change, remained compatible (strong selling point for Babel, has been compatible for 10 years). Reminder: in Babel, AE is used to encode both prefix (address) and next hop. 8: Two alternate encoding: "Toke", "Bastian+JCH". jch found Toke’s encoding too smart. Bastian+JCH encoding clearer and less sensitive to implementation errors but a tiny bit more verbose. 9: Draft exists. No tunneling. Intended status Experimental, no idea how to proceed from here. Discussion David: can you describe what backwards compatibility issues can arise with the Toke encoding? Théo: “if a route has no obvious next hop when it is received, the node should just drop the route” says the RFC. With this extension, we would accept the route. But if some implementation doesn’t respect this requirement and mishandles errors, there is a risk of the route being accepted by an implementation not supporting the extension. David: fair enough, my own implementation probably gets this wrong. Another question about Toke’s encoding when you mix v4 and v4o6. jch: In Toke’s encoding, we send either v4 or v4o6. Toke: I have no objection to using the other encoding. Margaret Cullen (Painless Security): It should be called v4-via-v6. Also, in a network where some routers route IPv4 but don’t have IPv4 addresses, there might be ICMP packets that a router is unable to send. This might or might not be a problem, but it’s worth mentioning. jch: agreed on both points. Babel Milestones & Charter ========================== Donald: [see slides: https://datatracker.ietf.org/meeting/108/materials/slides-108-babel-babel-milestones-charter] Reviewed work areas in current Charter, possible new work areas, and possible changes to current milestones. Barbara: Maybe we shouldn't add a milestone for IPv4 over IPv6 until the WG accepts the draft. Ronald in't Velt: I notice that the round trip time extension draft has expired. Is there any plan to revive draft-ietf-babel-rtt-extension? Donald: Feel free to send comments on the draft or questions to mailing-list, even if it is expired, or ask the authors about its status. Ronald: I am interested in it. Donald: Ok. If the original authors have moved on, you could even take over the draft yourself if useful/relevant. Wrap-Up ======= Donald: Thanks to everyone for attending. See you on the mailing list and at the next meeting in November which I think will probably also be virutal. Donald: One last chance if anyone wants to bring something up. [no response] Donald: OK, I'll end the meeting.