BABEL WG Minutes IETF 96 — Intercontinental Hotel, Berlin, Germany Thursday, July 21, 2016, 16:20 - 18:20 Potsdam II Room Notes taken by Margaret Cullen (Painless Security), edited by the Donald Eastlake Administrivia, Agenda Bashing ============================= Chairs: Russ White (LinkedIn), Donald Eastlake (Huawei) https://www.ietf.org/proceedings/96/slides/slides-96-babel-0.pdf Margaret Cullen agreed to take minutes. Barbara Stark agreed to be jabber scribe. No one suggested any changes in the agenda. Status, Review of Milestones ============================ There has been a call for WG adoption on the mailing list with several supporting messages and no opposition. No one in the room was opposed to adopting the RFC6126bis document as a WG item. Chairs directed the authors to republish as a WG draft. Juliusz: Thinks that first draft on “Related Drafts” list (diversity-routing) is out-of-scope for WG. Donald Eastlake agreed. It's the source-specific routing that is in scope. First Milestone item is completed (adopt base draft as WG document). Proposed Changes to BABEL Routing Specification =============================================== Juliusz Chroboczek https://www.ietf.org/proceedings/96/slides/slides-96-babel-2.pdf First slides discuss establishing traditions for group. Main points: work happens on mailing list, not face-to-face; if it isn’t implemented, it doesn't get in; BABEL user requirement focus. BABEL is not a clone of another routing protocol. RFC 6126 and 7557 should be one document. They were separate documents for historical reasons, but should be merged in the standards track version. BABEL version has been stable at version “2” for 4 or 5 years. Discussion of Security Slide: Margaret Cullen (Painless Security): Need to know what threats we are trying to defend against before we choose a mechanism. Juliusz: Need to know threats, look at mechanisms to handle those threats, and decide how complex you want the code to be. Justin Dean: MANET group has a TLV for security. Thinks we should change the BABEL packet format to allow use of the Manet TLVs. Juliusz: Willing to talk, but skeptical [details of response lost, sorry]. Don Eastlake: We will have difficulty with security area in getting mechanisms through if they only provide for static keying. Juliusz: Document under discussion has already been adopted as a WG document. Need to fix errata and integrate RFC 7557 — lots of meticulous editing, to make sure it reads as one document. Dave Taht: Does not think the term “primary metric” should be used for the loop avoidance metric. Since we support "secondary metrics" the most important of those should be called the primary metric and the metric currently called "primary" should be called "loop avoidance" Also, is there a use cases document? Do we need one? Alia Atlas (Juniper): No. There is an applicability document, which has some similarity to use case but not the same as a use cases document. Juliusz: Is use cases document needed in addition to applicability? Alia: No. We are here for things we can implement and play with. The WG was intentionally chartered with a minimal number of documents that do not lead to code. Russ White: Use cases can be useful, but don’t have to be drafts. We could put them on a wiki if that is useful. Justin Dean: Question on presentation: Should you worry about duplicate IDs? Justin would strongly suggest against detecting duplicate IDs in the WG. John Daddle (?): The use case thing did go through head when talking about security. Thinks you need to know use cases to work out security threats. Would strongly advocate for short set of use cases, to give the security ADs something to work with. Toke J: (Something missed). Doesn’t like the idea of using the MANET packet format, at all. Would not want to implement it. Alvaro Retana (Cisco): Slides talked about implementation and deployment experience before including a feature. Is that WG consensus? Don Eastlake: No. Not set by WG as an immutable rule. Just attempt to set culture. Juliusz: I used the word “tradition”. Would like to see the tradition live on in this working group. Benedict R: Regarding Justin’s comment on routing IDs — sometimes running on a embedded system without real-time clock. When you reboot multiple nodes (power outage), you can easily get a situation where you all get the same number. Denis O: Would like to note that running code tradition is not only a proposal for the WG, but has been the way things are done in the BABEL group. Would like to see it used here. Also, comments about security: would be great to receive comments on his recent post to the mailing list. Would like to get review/feedback. BABEL Information Model ======================= Barbara Stark, draft-stark-babel-information-model https://www.ietf.org/proceedings/96/slides/slides-96-babel-1.pdf Has been discussion of need for information model or Yang data model, so she thought she would throw something out there. Went through RFC to find everything that there is any way it could be included in an information model. Only base draft. Probably some of it should be thrown out. Did not attempt to separate state vs. configuration. Very little needed for configuration, and thinks minimal configuration should be support (none of which must be supported). Homenet for Hacker Boards ========================= David Taht https://www.ietf.org/proceedings/96/slides/slides-96-babel-3.pdf Dave: Can anyone run homenet on iOS? Tommy: Clarified he meant iOS (for iPhone), not IOS (for Cisco routers). Has run homenet and BABEL on an iPhone, but requires special in-house version. Denis: In older days, PPP was used for addressing, etc. on point-to-point connection. Could be considered? Dave T: Not going to see (very fast network technologies) on your home network any time soon. Juliusz: Says that we will never subnet beyond /64. Instead, we assign a /128 and do pure host routing. Lorenzo: Why not do a subnet length of /127, since that is a valid subnet length, and /126 is not? Alex Petrescu: Very happy to see IPv6 over USB discussion. (Lost part of discussion) Dave: Brought up questions from end of presentation. Is making BABEL work well on alternate link layers a goal of this WG? If so, which ones (USB, PowerLine, Thunderbolt, etc…) Barbara: Do you mean Non-Ethernet? Dave: Yes. ??: We already have IP over avian carriers :-) Barbara: Bluetooth and 6Lo don’t use Ethernet Mac, but some power line specs do? (Not X.10) Dave: We don’t measure quality of those layers at a level we need. Barbara: When we run them to a router, they look just like Ethernet. Extreme concern is that hopefully BABEL can tell they are not Ethernet (more delay, more loss), etc. Hopefully BABEL will not assume that anything with an Ethernet MAC is Ethernet. Alia: How do IP over different link layers is an Internet area. Important to decouple that from routing protocols, like BABEL. Keep problem decomposed and focus on what we need to do here. We did set complex metrics out-of-scope for the initial charter. Good discussions can be useful, but we have some basic work to get done first. If there are specific link layers that it is important for the Internet area to work on, they should go to the Internet Area. Juliusz: BABEL needs to distinguish two things: (1) link properties, loss, delay, etc. and (2) what MAC is in use. Need to distinguish between this WG (the standard BABEL protocol) and various BABEL implementations (BABELd, etc.). The implementations might do things (like detecting wireless links) that aren’t part of the BABEL Routing Protocol. Interesting work in the MANET space about metrics for different network types. Many people here have been doing interesting work on this problem, but we are not going to solve it here in the BABEL WG. Would like to see main document that says “we can work with any metric that has the following properties”, and informative or experimental work defining specific metrics (some of last part lost). Alia: One of the things that may be useful is to articulate what parts of the system are supposed to do what parts of the functionality? For example, one thing IP does is verify you can sent packets to a certain MTU. Discuss what sort of functionality does an IGP have? And are there other pieces that are needed? Dave: You can go and deploy BABEL on a bunch of boxes that cost less than $8 a piece today. Highly encourages people to go, experiment and find their own edge cases. Wrap-Up ======= Thanks to everyone for coming and participating. See you at the next meeting and on the mailing list.