IETF-76 MANET - November 12TH, 2009 Chairs: Joe Macker (via Jabber), Ian Chakeres (via Jabber) Proxy chairs: Justin Dean, Teco Boot Scribe: Emmanuel Baccelli Jabber scribe: Ulrich Herberg Minutes edited by Teco Boot, with audio recordings and notes from Ronald in't Velt ---------------------------------------------------------------------------- 0: Traditional Agenda Bashing: No comments. ---------------------------------------------------------------------------- 1: NHDP Way Forward Thomas Clausen presents draft-ietf-manet-nhdp status. WGLC on -01, few comments, one nit. NHDP adjusted for addresses, repeated on distinct, isolated interfaces. No need to include all addresses in hello message. Now document is ready to move to IESG. Discussion on finite state machine (FSM), to be included in draft. Or in separate document. An AD asks for this for std track RFCs. There is an I-D to be published soon with NHDP FSM. Ross Callon: Maybe better to talk to ADs that think that state machines are required. Personal opinion: state machines do not add much value. Thomas Clausen: We already started to work on state machines for NHDP and we think it is very useful. To be published in two weeks or so. Henning Rogge: State machines can be very useful to check if the other side is well behaving. Joe Macker: The State Machine is intended to be non-normative? Thomas Clausen: Yes, it will be non-normative document. Joe Macker: OK, then I think it is useful if it is non-normative. Teco Boot: I think we should extend the WGLC to give more time to review the document. Thomas Clausen: The WGLC has been going on for more than a month, and I believe we addressed all the comments. The document exists since 2006, we should put a limit on the time and move forward. We spent quite some time already and this is the second WGLC after the one last year. Also, we have multiple existing implementations. Teco Boot: There was not that much review output. Thomas Clausen: There was output, this is incorporated in the document. Ross Callon: Next reviews will not check details on MANET specific topics. We could get a directorate review if you wish, we could find reviewers. Thomas Clausen: If we involve someone that has not been involved in the MANET process so far we may incur more delays. We are essentially waiting for the MANET WG chairs to take action about this document, and forward it for AD reviewing. Teco Boot: Comments are always welcome. Thomas Clausen: Yes. But there are times and places for this. ---------------------------------------------------------------------------- 2: OLSR Thomas Clausen presents draft-ietf-manet-olsrv2 and draft-dearlove-olsrv2-metrics. Some delay, also with integrating metrics in base OLSR document. Delay caused by having first NHDP in good shape. When lots of addresses on routers, now this is simplified. And we have only one link between routers. This provides Faster calculation and makes implementing metrics more easy. Next step: incorporating metrics. Announcement of a machine document for OLSR. Henning Rogge: There could be state machines for routing metrics to check if the other side is working with the same routing metric. Thomas Clausen: It could be a good idea indeed. Teco Boot: I think it is important that we make sure implementations are compatible. What if we have negotiations for the use of metrics, and find out we are not compatible? Thomas Clausen: Oh yes. But what if we have a unitless value for a link, and we have additional values with a mechanism that negotiation. Interesting, we have to think about it, but it's not part of OLSRv2. Teco Boot: See it as early warning. Thomas Clausen - Announcement of a ETX draft. Strong demand for a document that describes ETX. ETX is used in large OLSR MANETs, e.g. funkfeuer. Funkfeuer: 400 routers, 1240 links, VoIP etc. with ETX. Also Athens, Leipzich, Berlin. Not to be ignored. Teco Boot: We could first have the document, and then ask the WG if we want this as default metric? Large community asks for this, I have also the experience we use OLSR cannot without. Same count for NHDP LQ. Thomas Clausen: This is an individual submission documenting operational experience. There are others not using ETX. We should have their mechanisms also. ETX may not be applicable to every L2. May not be a wise decision to have a default metric normative at all. Teco Boot: People posted: OLSR without ETX won't work. Thomas Clausen: I do not agree, there are L2 technologies that do not have benefits from ETX. Justin Dean: NHDP does not use ETX. Opinions on ETX is premature. We need to see the draft, and we need to make sure OLSRv2 and NHDP can use this metric, and then we'll see. Let's first request people to come forward with documents. Ulrich Herberg: I agree. Joe Macker: Its a good idea to begin to discuss the practices and motivations for various metrics approaches and get some of this written down. Henning Rogge: To Teco: IMHO, we should at least put a warning into OLSRv2 RFC that using "hopcount" is considered not very useful. And for this we need documents that describe how to do it in a better way. Thomas Clausen: We should warn people though, that this may not be the best metric for all deployments. Small networks with appropriate L2 can use hopcount. We may put a warning in. The important thing is, putting a dimensionless additive metric in. ---------------------------------------------------------------------------- 3: DYMO and SMF MIBs Justin Dean presents DYMO and SMF MIBs presentation. Default value clauses added. Questions: please on the list, so Robert Cole is able to answer. Thomas Clausen: What are the projected milestones for these documents? This is a general request to the WG chairs. When do we expect these documents to be stable? ---------------------------------------------------------------------------- 4: NHDP/OLSRv2 MIBs Ulrich Herberg presents NHDP/OLSRv2 MIBs presentation. Joe Macker: I would like to see some implementation experience with MIBs (related to Thomas' comment on MIBs milestones/maturity issue) anybody know of work? Henning Rogge: Comment to running code: we really need an efficient and small open source implementation of SNMP for small devices. NetSNMP isn't a very good way to do it. Thomas Clausen: Yes we need to think about it. This was also discussed in 6lowpan (ROLL?) (at the interim meeting in Geneva, comment by Adrian Farrell). Even a moderate constraint device has problems with running SNMP. Do we need MANET routers to be managed devices? Maybe yes, but not always. Maybe the MANET as a whole being managed as a single device and use a lightweight mechanism to get the info from a MANET device. Teco Boot: SNMP is complex, but maybe lightweight in bandwidth nevertheless? It just records a lot of data that is maybe not always needed. This may cause the heavy load. Ulrich Herberg: Yes this is a problem that is not MANET specific. Same question was asked in 6lowpan. Thomas Clausen: And also in ROLL interim Switzerland. Question is: do we want to manage each and every device? Or manage the whole MANET or ROLL network? Joe Macker: I also agree we need a lightweight mechanism. Hamid Mukhtar: There is work on this in 6lowpan, running SNMP over these constraint devices. The SNMP packets fit in 6lowpan packets, bandwidth is not an issue. Problem is agent size on the current devices. But what if it fits in future devices? (now first in MANET). Work on: how can we map the constraints, also for MANETs, how can we use SNMP? We can have end-to-end, we can have proxy. Analysis: existing SNMP agents cannot be reused as-is for the constraint devices. We need to look at the common requirements from MANET, 6lowpan, and ROLL. Teco Boot: Problem is getting data. Ulrich Herberg: Performance MIBs are not mandatory. Hamid Mukhtar: It is a placement issue, where do you want to place the MIB? Problem is not SNMP. Henning Rogge (supported by Thomas Clausen): The code size of a SNMP agent is a problem on small devices Joe Macker: Yes the size is an issue.. the patterns of collection/reporting is also an issue...different architectural issues at gateways, etc and with different node types... Thomas Clausen: Do 6lowpan people have operational experience with such MIBs? Hamid Mukhtar: Yes some guys in Korea have done that. There is lightweight IPv6 stack with NetSNMP over it. Maybe someone can also come up with lightweight SNMPv3 stack? Thomas Clausen: Can you cross-post stuff like this to MANET? Hamid Mukhtar: I was not aware MANET had same problems. Let's work together. I will forward info. Thomas Clausen: Ross what is your perspective about management framework for MANET nodes? And for ROLL and 6lowpan? Ross Callon: The goal is to be practical, and identify what is the most cost- effective way. In some case, we could have only some nodes that are involved in the management system, not all nodes, as long as the persons managing the network have their management data. We may want to bring an op AD to discuss this. Thomas Clausen: How do we go about and set this up since this is a cross WG issue, and we have similar constraints in ROLL and 6lowpan? Ross Callon: No problem to have separate MANET and ROLL to work on useful things. Thomas Clausen: Maybe better not come up multiple times with same issue. Ross Callon: I could sit down with other ADs. Teco Boot: But we need to continue with the current MIB documents. Thomas Clausen: Yes I agree. The question is rather: do we need something in addition to those. Joe Macker: We need some experience with the NOTIFICATION push style...Certainly some publications in sensor areas, etc There are multiple management patterns/styles don't preclude. Different environments, which manet address, (large platforms vs. embedded sensors, borders vs. other nodes) have different constraints. The REPORT MIB style provides some interesting bandwidth efficient style ---------------------------------------------------------------------------- 5: Report MIBs Ulrich Herberg presents Report MIBs presentation. Major change: history group. Typo on this version of presentation: not deleted "current history table". ---------------------------------------------------------------------------- 6: Security Extensions to NHDP, OLSRv2 Ulrich Herberg presents draft-herberg-manet-packetbb-sec presentation. Manet missing in draft title. Packetbb-sec is only syntax, no semantics Thomas Clausen: Not entirely true. Iana registry for codepoints, hash functions etc. So some semantics. Ulrich Herberg: First define security threats. Some are not addresses in the security draft, for example a router runs OLSR, but does not forward the packets. Replay: record messages and replay later or send out in another place. Te latter is wormhole attack. Thomas Clausen: The latter is wormhole attach, this is different than relay attack. Ulrich Herberg: Yes, but it uses the same mechanism. Thomas Clausen: The difference is perfect clock synchronization. Wormholes cannot be protected by timestamps. Ulrich Herberg: Key distribution is out of scope. The good guy has a valid key. We cannot protect against relay and replay attacks. Replay can be protected with timestamps / sequence numbers. This is optional in draft. Problems with having good clock. Henning Rogge: Do we want to remove signature TLVs? Or assume all zeros? Ulrich Herberg: We might have several signature TLVs. So remove first. Ulrich Herberg: Counterparts part of threats. Part of relay attacks. In NHDP we have no relay attacks. Key distribution is not addressed. Used algorithm is not suggested. There is no such that's fit everything. Way ahead, line up with DYMO. Line up with threats, see what matches, are we missing something. Thomas Clausen: Was this for the security threats draft? Ulrich Herberg: No, this for packetbb-sec. There is a new draft posted this morning: draft-herberg-manet-nhdp-sec-threats-00.txt Same for OLSR will be posted. Teco Boot: Is there a dependency between this draft and the protocols that will use packetbb? Ulrich Herberg: Packetbb-sec is containers for signatures and TLV. Nhdp-sec is how to handle this: how to sign and how to validate, this is not in the nhdp draft itself. Thomas Clausen: Each protocol may choose how it wants to validate messages. This draft is just the formats and syntax that can be used. We do not know if NHDP or DYMO would behave in a similar way. Maybe OLSR with MTR, preferring secure paths. Justin Dean: These drafts are good candidates to become RFCs. We need more people to look at this. Thomas Clausen: These drafts are in various "up-to-date" states. Packetbb-sec is in Good shape now. New versions of other documents are coming soon. Please do provide feedback! ---------------------------------------------------------------------------- 7: OLSR Interop Ulrich Herberg presents Report on 5th OLSR Interop Thomas Clausen: FunkFeuer network is 1240 links, 400 OSLR routers, there is a typo in the slide. Thomas Clausen: We will be able to announce all the details about the 2010 OLSR Interop very soon. Plans: Drexel University, close to NRL. Henning Rogge: We tested the software with 1100 real nodes... (Thomas: linked two OLSR domains). Hiroki Satoh: We have two more OLSRv2 implementations from Hitachi and from Niigata University. In October, we ran an interop test, with good result. Some problems because documents are confusing for us. Report follows. Ulrich Herberg: We will set up remote testing facility, with VPN. Also with a simulator. Henning Rogge: Is the OLSRv2 implementation available ? Hiroki Satoh: Niigata University code is public, and should be available soon. Report on this expected in a couple of weeks. Thomas Clausen: Can link be posted on list? Hiroki Satoh: I'll ask. ---------------------------------------------------------------------------- 8: DYMO readiness Justin Dean - DYMO WGLC readiness item: Thomas Clausen: Do we know about any implementations? I have not read the last version of the draft. Teco Boot: I asked this question to Ian a while ago. There seem to be several implementations. But I did not check them. On the status: there is the dependency on NHDP. Dependency is not that large. And DYMO has manet_id, but this is only DYMO. Thomas Clausen: It would be interesting for them to post experience on the list and come forward with such implementations. Manet_id is DYMO specific, this is not a problem. We would like to know if the packetbb-sec draft is applicable to DYMO or not as is, and if not, is there a way to converge? ---------------------------------------------------------------------------- 9: Open Discussion, WG Related Work & Announcements Ronald in't Velt: What is the status of SMF? Justin Dean: It was WGLC, there were few comments. No "big" issues, but no resolution yet. We are essentially waiting for NHDP since it depends on it. Thomas Clausen: I have not reviewed it and we agreed with Joe that we should finish NHDP first. Then we should probably have another WGLC for SMF. Joe Macker: SMF is waiting for NHDP WGLC resolution. I reported on the list about this. Justin is correct. Teco Boot: We discussed NHDP passed the WGLC. So we could go forward with SMF. Justin Dean: On a different note, I would like to encourage people to come forward with new drafts extending the current protocols, like the ETX draft and the security drafts, but hopefully others too, so that we can focus on them after we are done with the current WG items. I would like to encourage a greater participation from the people and not only from the core group of MANET.