MINUTES, MANET Working Group, IETF 120 -------------------------------------- Vancouver, British Columbia, Canada Friday, 26 July 2024, 09:30-10:30 (16:30-17:30 UTC) Room: Plaza A Chairs: Donald E. Eastlake 3rd, Don Fedyk, Ronald in 't Velt Area Director: Jim Guichard NOTE: time given below are original scheduled time, not time actually taken. Also, Don = Don Fedyk while Donald = Donald Eastlake. Chairs Administrivia (5 min) ---------------------------- [Chair Slides: https://datatracker.ietf.org/meeting/120/materials/slides-120-manet-chairs-slides-03] [Slide 1] Don Fedyk (Chair): Hello, Welcome. This is the MANET WG. We have a one-hour session so we will get going. [Slide 2] This just gives some information on the WG. We have Ron in the room and Donald and Jim Guichard, our AD, on line. [Slide 3] Draw your attention to the Note Well slide. We have been told to leave this up for a bit so you can read this. If you have any questions, ask. [Slide 4] We also want to draw your attention to the Note Really Well slide about behavior and anti-harassment. We try to maintain a good working environment. [Slide 5] Some more notes. Be sure to log into the Datatracker. This is a one-hour session so ANIMA will be taking over this room at 10:30. If you are in the room you can use the light client. Be sure to use the client to get in the queue. [Slide 6] We have a fairly short agenda. We have the agenda bashing. We will be going over the milestones. We have a talk on an OLSRv2 tweak by Christopher Dearlove and we have a talk on IPv6 Addresses for Ad Hoc Networks by Fred Templin. Then we have reserved some time for Charter revision discussions and wrap up. Draft Status, Milestones (5 Min) Chairs --------------------------------------- [Slide 7] Ron in 't Velt [Chair}: We have been making progress on credit window DLEP drafts. Two of these drafts are through IETF Last Call and two have just entered IETF Last call. Then there are a bunch of Directorate reviews going on. They are in the pipeline. Then review by the ADs and we expect comments. I am the document Shepherd for these drafts. [Slide 8] Then we have some drafts by Henning Rogge about physical layer based extensions that were adopted by the WG about 5 months ago. We really, REALLY could use some WG comments on these drafts. Don has just commented on one of them. There are also two other related drafts, one by David Lou that was presented last year in San Francisco. The draft has been updated but he didn't respond when we issued the call for presentations for this meeting. The AODVv2 draft was presented in Brisbane by Charlie Perkins. We will talk about that a bit more when we get to the Charter discussions. Jim Guichard [AD]: I just wanted to clarify for the WG that the four documents I have right now I want to move forward as a block. That's why I have not moved the two drafts that have gone through IETF Last Call. That way the IESG can consider them all together. Ron: That's all right. It is what we would prefer. Don: If we have any note takers, they should contribute to the notes. Next up on the agenda is Christopher Dearlove. Chris, you have the floor. Responsive Use of OLSRv2, Christopher Dearlove (10 min) ------------------------------------------------------- [Slides: https://datatracker.ietf.org/meeting/120/materials/slides-120-manet-a-responsive-olsrv2-proposal-00] [Slide 1] Christopher Dearlove: This is a presentation on a short draft to fill a small gap in OLSRv2. [Slide 2] Will talk about one scenario but this isn't limited to that case. [Slide 3] The issue is networks that do not change often. One approach is to back off and send messages less often. This draft is motivated by the more extreme case of sending no messages. [Slide 4] OLSRv2 has the option of only sending responsive messages. Normally those would be in addition to periodic messages, and you can always set the message interval to some very large value to get this effect. [Slide 5: Local Effects] Two effects we are talking about: a new router arrives, or a router leaves, but everything else is static. If a router arrives, the HELLO message exchange will update all neighbors and two-hop neighbors. If a router departs, it might send a HELLO with no neighbors before leaving but if it does not, there can be a problem. And we can't rely on the data link layer noticing. [Slide 6: Remote Effects] We might still need periodic HELLO messages but what about TC messages. (These are the main barrier to scaling.) It would be nice to top sending them. This has two effects: Routers accumulate obsolete information, but this does no harm as it is never used, New routers send responsive TC messages so remote routers learn about them, but that does not work in reverse. That's the reason for this draft: Remote routers will learn about them, but they will not learn about remote routers. [Slide 7] On solution is some sort of local database exchange and that is referenced in the draft, but it is not backwards compatible. So, this is not part of this draft. The solution within the scope of using existing messages is that a remote router sends a TC message when a new router appears. It knows this because it is adding a remote router tuple. The draft has a suggestion for when it MAY, SHOULD, or SHALL do this. MAY is true currently. SHOULD is if you are heavily relying on responsive messages. SHALL [MUST] if you are fully following the approach of no periodic TC messages. This is the normative part of the proposal. [Slide 8] Any Questions? Juliusz Chroboczek: The idea of having reactive features in a proactive protocol is a very old idea. The oldest instance I know of are the reactive extensions to RIP. That was deprecated because it did not work very well. Many people have tried to do this with the result that, if you can get it to work, it does not work very well. The problem is that if you are being reactive in a proactive protocol, there is some point where you are doing path discovery, TC message in your case, and you find that the amount of traffic you have is more that you would have if you were purely proactive. I am not saying you will not succeed, what I am saying is that if you do succeed, I will learn a lot. So, I encourage you to try it out but without evidence that it can be made to work, then I'm not sure a maintenance group should be working on this. Christopher: That's a fair comment. Unfortunately, my position does not allow me to try this out. I am a retired person. Juliusz: Then you should be in a great position to try this. Christopher: I am trying to negotiate access to the code I wrote for my former employer. Juliusz: I understand. If you get this to work, then I am going to steal your ideas for BABEL. You should take this as a complement. Christopher: OK. Any more comments? [no one spoke up] Don Fedyk: Is that it? Christopher: I would say the options are that the WG could adopt this, or it could say, "No, we don't need it." Should be discussed on the list. Don: We'd like to take it to the list for comments, then we can consider the next action based on the comments. Ron: This is fully covered by the current Charter as OLSR maintenance and extensions, so that would not be a problem. Stuart Card: Responding to the comment on putting reactive features in a proactive protocol. In my experience, that will be sensitive to the rate of topological change, particularly joins and splits. [Two or more networks joining to become one or one network splitting into two or more networks.] Christopher: The case I was thinking of does not includes joins and splits. IPv6 Addresses for Ad Hoc Networks, Fred Templin (10 min) --------------------------------------------------------- [Slides: https://datatracker.ietf.org/meeting/120/materials/slides-120-manet-ipv6-addresses-for-ad-hoc-networks-01] [Slide 1 Intro.] Fred Templin: Imagine we were all transported to the Northwest territories. There are no cell towers for hundreds of miles, but we should be able to form a little ad hoc network among ourselves, between our devices. [Slide 2] That would look like this. ... The question is what kind of addresses do we use? RFC 5889 recommends GUAs/ULAs but isolated Ad-hoc network coordinated provisioning is often not possible. LLAs are limited to single link but we have a multilink. We need MLAs, Multilink Local Addresses: high assurance of uniqueness, self-generated or coordinated with a registration authority, the uniqueness must persist across network partitions/mergers, must support local multihop forwarding over multiple links. This MLAs are assigned to ad-hoc network interfaces as /128s. [Slide 3] MLAs are well formed, valid IPv6 addresses. ... They apply equally to static networks that are not particularly mobile or wireless which is why this draft is targeted for the 6MAN WG, not the MANET WG. [Slide 4 Skipped] [Slide 5] There are three type of MLAs we are looking at. This is "Type 1", IPv6 prefix 2001:TBD::/28 followed by 100 random bits so probability of conflict is low. Can be self-generated, should not appear in global name service. [Slide 6] "Type 2" ORCHID (Overlay Routing Cryptographic Host IDentifier) [RFC7343]. 2001:20::/28 followed by a 4-bit ORCHID Generating Algorithm ID followed by 96-bit Encode_96(Hash) of Context ID. Can be self-generated but have fewer pseudo-random bits. Can serve as a node identified and may appear in global name service. [Slide 7] "Type 3" Hierarchical Host Identity Tag (HHIT) or Drip Identity Tag [RFC9374]. 2001:30::/28 followed by a 28-bit hierarchy ID followed by 8-bit HHIT Suite ID followed by 64-bit ORCHID hash. Must be registered with a registration authority before use. Nodes can assert registered Type 3's as unique addresses. Can serve as a node identifier and appear in global name service. Supports attestation of a node's claimed address and address-based authentication. [Slide 8] What about the old fec0::/10 site local addresses? These were deprecated by RFC 3879 and even today, 20 years later, many implementations refuse to recognize these addresses. MLAs could replace site local. [Slide 9] ... [duplicates part of Slide 3] [Slide 10] What about Ad-hoc networks connected to the Internet? Say intermittent or permanent connection. You can see on this slide, the node at the bottom right, say a StarLink ground station, provides internet connectivity. It can assign addresses to all the other nodes using SLAAC or DHCPv6 or other means. The Ad Hoc network nodes assign GUAs/ULAs [Globally Unique Addresses / Unique Local Addresses) to their Overlay Multilink Network (OMNI) Interface, not to their ad hoc interface. Ad Hoc networks access Internet services via OMNI encapsulation (IPv6 address selection rules apply). [Slide 11: Next steps for this draft] https://datatracker.ietf.org/doc/draft-templin-6man-mla It is currently targeted for the 6MAN WG since it applies for all Ad Hoc network types, not just MANETs. Supporting comments on the 6MAN list welcome but comments on the MANET list would be good also There are also two other related drafts listed here. Fred: The main focus to take away from this draft is the need for Multilink Local Addresses and the three types we are considering. Don: Any questions? Stuart Card: The work we did in the DRIP WG on the Hierarchical Host Identity Tag was primarily intended for a different use case but we always anticipated it might have wider application. We would be glad to work with you and support this other use case. Just want to make sure that each identifier type gets used for the appropriate use cases. The Hierarchical HIT is very different from the HIT. The HIT you just generate yourself. The Hierarchical HIT was intended for the case where there is a registration process. You can only prove the validity of a Hierarchical HIT if you have an endorsement that proves you are in a branch of the tree, Stuart: A long, long time ago I was involved with a BoF that persisted for several months as an email group that was looking for a router ID. This would identify the router itself as opposed to IP addresses that identify interfaces of the router. How do you pick one to use for the Router ID? We decided that a HIT was appropriate because it does not imply any topological aggregation. That is the key thing to keep in mind: When does topological aggregation matter and when doesn't it matter. When are you looking for a locator and when are you looking for an identifier. Fred: In this case, with the ad hoc network in the wilderness, we are strictly identifier based and that /128 can act as both a Router ID and an interface identifier at the same time. Juliusz Chroboczek: I am ashamed to say I have not read the draft, but could you say a few words about why those addresses are not ULAs? I am looking at the properties on slide 3 and they just look like ULAs. Fred: The thing about ULAs is that they have a subnet ID field to provide a common subnet for nodes that assign addresses from the ULA prefix. Juliusz: If I may interrupt, if you are doing a mesh network, you can just consider that all 121 bits are the node identifier. Consider the ULA space as a /7, draw 121 random bits, and the MANET protocols are able to deal without any explicit prefixes. It is not clear to me why the MLA space is useful here. Not that I am opposed, I am just trying to understand. Fred: Right. The problem with using ULAs like that is that it conflicts with the use of ULAs that would be more structured. By using these MLAs with a well-formed IPv6 prefix, we don't get to the problem of conflictions with structured use of ULAs that we might get in another network. Juliusz: Are you worried about collisions or about debugging issues? Fred: I'm worried about nodes that would think they are in a ULA subnet when they are not. Juliusz: OK. Thank you. X: A problem might be that you are using ULAs in the overlay network. If you use them in the radio network how does the device know what's what? Fred: Actually, you assign any particular address to the ad hoc interface or the overlay interface and the provides the context to interpret the address, Don: Thank you, Fred. If there's any other comments, take them to the list and look at the 6MAN list also. Next is the Charter discussion. I'll hand it over to Ronald. New MANET Charter Discussions, Chairs (25 min) ---------------------------------------------- [Slides: https://datatracker.ietf.org/meeting/120/materials/slides-120-manet-a-charter-text-proposal-00] [Slide 1] Ronald in 't velt [Co-Chair]: Yes, thank you. I put to together a proposed Charter text. Actually, I should say Donald put together a proposed Charter text and posted it to the list at the beginning of June. Unfortunately, there were no reactions. (I didn't comment either because I was otherwise engaged.) I have now done my homework. I put every paragraph on a different slide and I will walk you through them. The first few should be noncontentious, but you never know. then we get to some things that might be contentious or at least deserve discussion and I hope that we can start the discussion here and obviously it will have to be continued on the mailing list. So, here we go. [Slide 2] This is a bit of introductory text. I have removed a few little bits that I think were left over from the formation of the WG and we silently passed over our 25th anniversary last year. So, this is just what we are doing. Standardization of MANET but also maintenance and extension for existing work that has already been done. This includes the routing protocol developed in this group, OLSRv2, as well as BABEL because the intention is that as soon as the new Charter is in effect BABEL WG will be closed down and BABEL activities will be continued in the MANET WG. Ron: By the way, if at any point there is something you want to react to, just raise your hand. [Slide 3] This is some old text I retained mainly to highlight the point that it doesn't have to be all wireless links. BABEL is defined to support both wired and wireless links. DLEP is a protocol between routers and attached radios. For some reason the radios got named modems; I'm sure that Christopher will remember the discussion. The idea is that the routers can make better decisions if they have link information available from these radios or modems. At the beginning of the last Charter, DLEP was a work in progress but since then it has been done and several extensions to DLEP have been in defined. So, we are now in maintenance mode for DLEP. Some more extensions are in process and there might be more coming in the future; however, I cannot think of another 100 DLEP extensions, it should be far less. So, it should peter out at some point. [Slide 4] Now here is where we can hopefully have some discussion. At the last meeting Charlie Perkins presented AODVv2 again. This has been a work that was originally done in this group. At some point it was abandoned, around 2016. One of the reasons it was abandoned was that it was holding up the rechartering back then. It then progressed as an AD sponsored draft for a number of years and now the question is do we need to bring it back? Is there energy in the group to work on this? Do we want this? Do we want the reactive protocol at all? Or do we just leave it be? Rick: I think AODVv2 (AODV X) is a fascinating concept. I would love to see it make progress. I think the group needs to ask some questions. Does it need a document editor because Charlie is making very slow progress with it. He is quite retired now. Some fresh blood to make progress on it would be important because the danger is that it may go in the Charter and in two yearsÕ time we will have made little progress. I would like to see it done. I think it needs inputs from the WG as a whole to make it happen or it shouldn't be in the Charter. Ron: I tend to agree. There is no point in putting something on the Charter that no one wants to work on. Christopher Dearlove: I would say the draft, in the charter, the main advantage of reactive protocols is low overhead. I would say "potentially low overhead". It does depend entirely on the scenario whether or not it is. With regard to the last comment, yes, I haven't looked at Charlie's latest AODV draft but certainly when it fell off working group status, there were three or four questions that I think would need to be addressed before you adopted it as a working group item because otherwise we would be back around in the same circles again. Ron: Yes. Juliusz: I'm not going to speak to whether we should adopt it, I'm just going to object to one sentence in the proposed charter. You say the main advantage of reactive solutions is low protocol overhead. Okay, so there's a discussion on that. There is a lot of literature on the subject and the results are inconclusive. I donÕt think that the charter should be making assertions in this discussion. I would suggest simply removing this sentence. It does not belong in the Charter. And, I suppose, if you disagree, we could say something like the "claimed advantage" of reactive solutions. Ron: Fair enough. I will remove it. Stuart: IÕm deeply divided here. I am interested in hybrid routing protocols. I don't know if anybody here remembers Zygmat Haas's zone routing protocol ZRP, things like that. Rick and I have had discussions about casting all of this is a more formal probabilistic framework. And if you going to have a hybrid routing protocol you have to first have a reactive and a proactive protocol. So, I'm in favor of keeping reactive in play here in the WG. On the flip side, I am completely consumed with the DRIP [DRIP WG] work and do not know when I would be able to dedicate any energy to such an endeavor. Ron: OK, so you are not making a recommendation either way. Christopher: I think the basic mechanism of AODV is well understood and actually you don't need a fully defined AODV before you could create a hybrid protocol as an extension of OLSRv2. It's an area related to my previous employment so there's a limited amount I can say about it. We played around with it on paper but not necessarily any further. There are one or two questions we encountered but I think the key question as indicated is effort because without effort you are going nowhere. Abdussalam Baryun: I just wanted to say that I am interested in this work, and I am willing to give effort with the authors. Thank you. Ron: Good to hear, thank you. [Slide 5] Ron: This is about multicast. We also discussed this several times in previous meetings. My opinion, and at least Rick has made statements to the same effect, is that if you have a homogeneous layer 2 or layer 1 technology, you can do things to support multicast below the IP layer and then there is no real point in doing another layer of multicast at the IP layer. However, this changes if you are using several of these layer 2 / layer 1 technologies in combination or if you are federating different MANETs with different proprietary technologies. Then you need some kind over overlay routing protocol to make this work. I'm not sure we are going to define a standards track protocol here but I think we should pursue this a little bit further and maybe do a requirements document or something. Juliusz: First, I strongly support this and if I had infinite free time I would start working on it straight away. Unfortunately, I cannot estimate when I will have infinite free time. To speak about the actual text, I think that the first sentence is too weak. You're saying that the problem is the amount of churn but even in a static network PIM is not going to work because it uses RPF and RPF doesn't work in mesh. I'll send you on the list suggested rewording. I think that here we can be much stronger. Ron: Yes. Text contributions are most welcome. [Slide 6] Ron: This is material about coordination with other working groups. It is copied almost verbatim from the current Charter. Stuart: In my work, multicast over heterogeneous media is critically necessary and, if there were at least three of me, then at least one of us would be working on this. There's not. Rick: Two quick points. The first point, echoing the previous comments of "I'd love to work on this but", I think it is important for those of us who are long time participants in MANET to go out and find more resources to help burn these cycles because we understand that these topics are important. My second point was on the multicast piece. Ronald, I think you touched on something quite elegant there with the heterogeneous multicast where the homogeneous islands of link level technology are capable of handling multicast natively, there may be a middle way through this. Rather than trying to build another multicast overlay network, more of a gatewaying protocol between island of multicast capable MANETs in a sort of BIERish [BIER WG] way, that might be more achievable than starting with sparse mod or dense mode and trying to build something new, which seems like a lot of work and quite researchy. Ron: Yes. I can see that avoiding looping could be an issue but let's not have that discussion now. [Slide 7] Ron: This is the last slide. Here I enumerated a number of potential work items without claiming completeness. So, if you have other ideas here, let us know. Juliusz: I'm not sure how you define extensions, but if you are going to put it explicitly in the Charter, I would feel more comfortable if, whenever you say extensions, you would add "and maintenance". Just in case someone tells us "ThatÕs not an extension". Ron: Fair enough. Sure. Abdussalam: Regarding multicast solutions for MANET, the point there says itÕs based on heterogeneous wireless technology. But what if it is not based on heterogeneous wireless. We need at least a solution or a standard we agree on even if itÕs not heterogeneous. We have for MANET multicast protocol. LetÕs say either if it's proactive or reactive at least we have a multicast protocol for MANET. But here if we just say based on heterogeneous, maybe it is not a good start. Just say we need a solution for multicast, and we start with it. Also, regarding the other items or even how to cooperate with other working groups, I see that it is better that MANET is an old WG and the other groups were maybe started by cooperating with this group. So, it is usually case-by-case or let's say I see some working groups they start with a case study or with a topology and they start working on a protocol that is related to our protocol. So, we can work similarly case-by-case. If you say as here multicast based on heterogeneous, we can say a protocol or such protocol which is based on such case or such application. Abdussalam: And then we can have that kind of protocol or the authors or effort on application or who are effort on that case. They can cooperate with the other working groups. But if we just say generally, now let's say we have OLSR version 2, okay, it's a generally it can be worked on many cases, but why don't we have some kind of group checking it, how it works on let's say, the satellite case. And we have already a group in IETF working on a satellite protocols. And also, we have other cases say like the ROLL [ROLL WG], you know, or the low, low-power, so that's another special case for... Ron: Sorry to interrupt you but we're almost out of time. I hear what you are saying. Can you take this to the list so that we remember all the points that you made. I won't have time to answer all of these things now. If you can send an email then, at least, we will have all of your points on the mailing list. Wrap-up, Chairs (5 Min) ----------------------- Ron: We have less than three minutes left. [To Don] Is there anything you want to say to wrap up? Don: Maybe just if Jim has anything to say, we have a couple of minutes. A reminder that ANIMA is here right after us, so we are going to clear out pretty quickly. Jim: Yeah, Don, I don't really have anything more to add other than what I have already stated. Which is, in order to recharter, there need to be, obviously, technical work to do but then you need the bodies to actually do the work. What I've heard in the session so far is that there are some people who are willing but pretty much everybody is unable. So, that's a little discouraging. But let's just work on the Charter to get it to where you think it should be. But then we need make a decision as to whether the recharter is worthwhile if there is literally no one to work on those items. On Charlie's work, I mean, I'm not going to recharter a working group to work on a document that only one person is actually working on. There are other avenues for that, that Charlie needs to continue to work. Obviously if there are other people in the working group that are actually going to commit to the work, then that's a different conversation. Don: Right. Okay Well, I think that pretty well does it. We've got ANIMA people coming in the room, and we're going to be clearing out. So, thank you all very much. If there's anything else, please take it to the list. Have a good day. Ron: Thank you, bye. Donald: Thank you