Meeting : IETF106 Wednesday November 20th, 2019 Venue : Raffles City Convention Centre, Singapore Time : 10:00 to 12:00, during Morning session I (120 minutes) Location : https://datatracker.ietf.org/meeting/106/floor-plan?room=hullet#raffles-city-convention-center Chairs : Erik Nordmark Rick Taylor Responsible AD : DEBORAH BRUNGARD Live minutes : https://etherpad.ietf.org/p/notes-ietf-106-raw Live feeds : https://datatracker.ietf.org/meeting/agenda/ Other URLs : https://tools.ietf.org/wg/raw : https://datatracker.ietf.org/wg/raw : https://www.ietf.org/mailman/listinfo/raw on Wednesday November 20th from 10:00 to 12:00, during Morning session I in room Canning Minutes ---------- Ethan Grossman (Dominique Barthel as a backup) Intro, why this BoF 5mn The chairs, unique attendance (wireless) new problems (RAW) Use cases and reqs 10mn Carlos on stage, show unique use cases, demonstrate usefulness draft-bernardos-raw-use-cases Select the particular use case of indistrial automation as illustrat Uma Chunduri: Layer 1,2 shown - is there layer 3 work? Erik Nordmark: Multiple layers of technologies. Uma Chunduri: 6TiSCH is layer 1 EN: Is about layer 3, like DetNet, which presumes different data plane (layer 2) technologies. Problem Statement 35mn Pascal on stage, what needs to be solved, position vs. other WGs draft-pthubert-raw-problem-statement Janos: Discussion now or later? EN: Now is fine, but if open-ended save til later. IEEE Status 10mn Ganesh, progress and unique features in .ax and .eb towards RAW Juan Carlos Zuniga: Is this MAC going to work on different PHYs? Ganesh: From what I know it is for the 5-7 GHz bands Janos Farkas: Reiterate that in IEEE there is integration of TSN into 802.11, resulting in subnet technology that could be used for DetNet. Technologies draft 5mn Pascal generalizing over abstracted technology for L3 Niels at the mike on LDACS Pascal Thubert: There are additional slides on LDACS if you want more info. draft-thubert-raw-technologies Other drafts 5mn chairs, demonstrate interest and activity Charter Discussion 35mn Erik and/or Rick holding the pen, edit and drive the discussion Demonstrate tractable problem, actual IETF topics, no overlap Lou Berger: Path selection: aspects: fast forwarding decision based on lower layer state. We do lots of things like this in IETF, can't say we haven't done it. Like ECMP. In DetNet do PREOf, nailed up version. In DetNet definition, allows for network coding, forwarding layer based on link probability of delivery. Again long history of this. EN: Agree. Rick Taylor: DetNet has lots of work to do, is DetNet really interested in working on this? Purpose today is to determine if there is enough interest in this work in other WGs, e.g. CCAMP. LB: DetNet wrapping up data plane drafts, including def of PREOF. Do we have time now, as we are rechartering? Probably we will have time, but is it the right place is another question. Definitely have people sitting around who are not interested in the particular technology e.g. optical. Pascal: We expect to draw a wireless crowd to RAW - so far, Detnet hasn't gotten people interested in this [wireless?] so far. So might be split between wired and wireless sections. RAW can focus on one particular thing, but interact with other WGs e.g. DetNet. Pascal has tried to get DetNet interested in wireless, but could not get support. Can see what the focus in the Detnet room is, different than radio people. Lou: With many technologies, do we want to create specific methods for specific envitornments, or a toolkit that is more general? Identify what is more common so can apply internet technologies. Want more commonalities, not less. Is this too much of a taylored solution for wireless, or better to have common mechanisms between technologies? Lou: In DetNet, have always had interest in wireless, but had to do basics first. Now have done last call on data plane, ready to take on more if this is the right place. Janos: What is the ultimate goal? To provide deterministic networking across wired and wireless, agree we need to identfy gaps and fix them. Service sublayer, then forwarding sublayer, maybe this is a lower layer, so could do in DetNet from top down perspective. Up to now, sounded like wireless work proposed for DetNet was an extension to 6TiSCH. DLEP can help with APIs. 802 subnet improving to the point where could treat as a data plane from DetNet perspective (?) Could do this work in DetNet and other groups such as CCAMP. Rick: How many people intersted in going to DetNet to solve this? (not too many?) Michael Richardson Could be because DetNet working on differnt things now, so isn't the time to go there. Erik: Where to do new work items, which depend on radio people? Up to ADs to decide. Stuart Card: IETF has historically left airborn networking to airborn communications companies who only know how to do links. Need to address Ning Zong: Would like to see architecture in more details. How is this based on an L3 abstraction? Hard to agree on this charter w/o this info. EN: Would have liked this in today's preso, as opposed to in future work? Ning Zong: ?? Lou: Not going to take on the work of radio guys, DLEP enhancmenet in MANET, Architecture, framework, how to apply to a given environment. Whether do in new WG or as collaborative between existing WGs. Just need to decide where. Rick: requesting a home to do this work. Tim Costello: Good work. Regarding other WGs, maybe could make other WGs more friendly to wireless? Rick: ?? Janos: Don't want to see new architecture coming out of thisk should be in DetNet. Fatai Zhang: PCE said to be out of scope on slide, but how can you do it without affecting that. Should make PCE fast. Also comment on pb statement, comparison Rick: REsponsibility is not defined here, just saying need separate time scale for path selection - where it happens is to be determined. Erik: Time scale is related to link type - if have unreliable link at varying speeds, maybe could not react fast enough. David Black: Work items: first and last two items relation to DetNet: First is describing how it is different than DetNet, can be independent. But redundancy may overlap with DetNet PREOF. Don't want to charter separately from DetNet version, or separate for wired wireless. Erik: Dont want differenet arch, just add missing peices for wireless to DetNet. Separate group to do that? Lou: Regarding architecture, agree so far just concepts. Have one solution today, but maybe #4 (PAREO) could be done in DetNet, bring a proposal. Regarding OAM, should do in DetNet to avoid duplication. DetNet is not "wired" it is any medium. Rick: Our intent is to work with other groups. If we show up at DetNet with 9 drafts tomorrow with a group of people who want to work on them. Lou: #3, 4, complementary to what is being done in Detnet. Others can be accommodated. Want commonality, that is what IP is all aout. Not helpful to have multiple solutions. Jari Arkko: What is market demand for this work? would like to see more on this. Clearly demand for low latency and so on, but maybe this is mostly layer 2 work? Are we creating a system on top of this? Choice between work done at L2 vs L3. If bridging multiple standards need IP. Layer 2 know more about state of medium, not so much at L3. So maybe this is less IETF work, and leave some of this to radio manufacturers? Rick: Question is whether this is an IETF project? Building across multiple technologies. So there is a problem for IETF there. Jari Arkko: Is there a market for this? Rick: I work for Airbus, our industry needs this. Thomas Graeupl: Hard to find people at IETF to talk to about standards for aeronautics communications. Having a focussed WG would help. Pascal: Must have multi-hop routing at L3, can't rely on L2 meshes. .. ?? Neils: Need to do good channel measurements, important for UNIs. RAW WG Good opportunity to bring in new people to IETF to do channel measurements. Uma Chunduri: Is this a new data plane? Not DetNet. Erik: Hum for yes or no: Problem statement and work items are sufficiently clear. (Seems most feel is clear, but maybe a few not clear). Solve in IETF: Agree should be in IETF. Support for a separate working group or not: (about 50/50??) but a lot of abstention Deborah Brungard: If do WG for only first 2 work items, informational, do poll again: For/against was maybe 80 vs 20. Who in this room would Maybe if only do those two, isn't interesting enough? Pascal: PSE is an extension to DetNet. What if DetNet delegates this work to RAW? Erik: If start work on this, whether in RAW or other groups, how many will work on this? Show of hands. Result: about 12. Willing to review this work: Maybe 30 or so. There is a RAW list, please use it if you have further comments. Adjourn.