Lwig Status PagesLight-Weight Implementation Guidance (Active WG)
Int Area: Suresh Krishnan, Terry Manderson | 2011-Mar-27 —Chairs:
IETF-99 lwig minutes
Session 2017-07-20 1810-1910: Athens/Barcelona - Audio stream - lwig chatroom
LWIG @ IETF 99 - July 20th, 2017 Note takers: Mohit Sethi New Note Well 0. Agenda bashing 5 min Some draft needs more work before pushing to IESG. Suresh, AD: Energy Efficiency. Requested IoT and Int review. AD eval follows. Hope it will be with IESG in a month or month and half. 1. Minimal ESP Daniel Migault ESP likely to be used by small devices. Document has received significant feedback. Authors believe doucment ready for WG adoption. Changes sinces 04: - Clarified purpose - Not designing another version of ESP - Clarified text on SPI - How to avoid generating random SPI. Randomness is still needed. - Padding was corrected and mentioned as mandatory. - Next header - ICV clarification - Known implementations @Hannes: Can you tell something more about implementations. Which of the IoT OS implement that @Tobias: Shahid Raza implementation in Contiki. LMU @Hannes: We should not do that work. IPSec does not provide benefits. Add fragmentation. Doesn't do us a service. @Tero: 15.4 peer wise. Cannot do TLS. Don't need TLS. For constrained devices you pick one. Use IPSec everywhere and don't use anything else. @Hannes: Are you going to use this. @Behcet: Out of Scope @Hannes: Standardizing something. Done something. Guidance. @Suresh: Early independant draft. Why do you see a need for this. We are figuring out if somebody else has a need for this. Bit early. @Tero: Not standardizing something new. What kind of tricks you can do. Same thing we did with minimal IKE. NAT traversal. In most minimal cases. @Hannes: Bad idea to standardize minimal IKE. Stripped out randome things. Right thing to do in some deployments and not in other. Feels arbitary. @Tero: 15.9 does not need NAT traversal and uses minimal IKE. Using this is better than proprietary. @Zhen: Hannes you supported. Work is not normative. @Mohit: RFC7942 Improving awareness about implementation @Zhen: You want to continue have in LWIG? Do we ask for call for adoption? If you think we need to solve this in LWIG group. Solve yes: More than 10 Solve no: Few (One) 2. TCP over Constrained Nodes Carles Gomez Montenegro CoAP originally designed over UDP. CoAP over TCP in progress. Other application protocols such as MQTT, XMPP for TCP. TCP is being and will be used in constrained node networks. Status: Updated in IETF 97,98 and today is last update 03. Details on RIOT and OpenWSN implementation MSS: Desirable to limit the MTU to 1280 bytes Delayed Acknowledgements- Disabling delayed ACKs is recommended @Markku Kojo: Confusing if traffic is transactional. Server is not responding slowly. ACKs come with response. That is not transactional if data is not coming back from server. Another updated to annex of document: RIOT TCP implementation. @Rahul: Differ. Even in case uIP. Impossible for application to generate same seq number. @Carles: This case the app needs to have reliability. @Rahul: Current uIP. Every connection has one buffer. @Carles: Need to verify versions. OpenWSN implementation : no retransmission WG adoption. Authors think it is stable. Getting more complete. @Abhijan: Slow start. Not utilizing b/w. Work on limited transmit. RFC3402? Improving TCP performance. Working on something similar. Considering this dimension? @Zhen: Did you read draft @Abhijan: yes I have. @Carles: yes, we can consider this. If maybe TCP fast open maybe good idea? We can revise the document. @Abhijan: That will be good. @Michael: TCPM co-chair. I suggest adopt. Work is needed. Heading in right direction. Milestone may not realisitic. Read this draft: 10 Supporting this: About 20 Not supporting: None 3. Neighbor Management Policy and Implementation Consideration for 6LowPAN - Rahul A. Jadhav -Presentation: Node density can be anything. Cache size is small. How do you prioritize enteries here. Ensure proper routing such that network paths don't keep on changing. 50-100 nodes. Neighbor table size is 10. You need to have a neighbor managmenet policy. RIOT is FCFS DAO storm. Neighbor cache is full of parent enteries. 100 of nodes in one deployment. RPL is example. PANA client. Same problem exists in 6tisch. Neighbor management operations: Insertation Eviction- Can evict non preferred parent. Use this intelligence. Reinforcement Clearing unused NCEs. Propose a reservation based policy. All this is reactive. However this has limitations. Child has lost state information. No way in RPL. Clearly out of LWIG. Like to bring it to ROLL -DISCUSSION @Michael Richardson: If it was only implementaiton advise. Maybe this belongs in neighbor discovery. RPL networks without actually doing router advertisements. Maybe put it in both DIO. There is protocol action. @Carsten: If hosts in ND. IF routers then in ND or routing protocol. @Michael: Needed by both. @Carsten: Then it is ND. @Zhen: Do you have comments on Reactive. For proactive which is right working group. @Carsten: Even if 6lo has done work, it is good to have work how to do it. Useful document to have. Needs comments to 6lo-6775-update document immediately. @Michael: Agree. All of advise here. @Rahul: Keep Read: 5 Adopt this document in LWIG: About 8-9 Not adopt this document: None 4. Minimal g-IKEv2 by Tobias Guggemos 5min Group IKE is going to be standardized soon No document Follow up on minimal IKE. We did find useful group key management @Hannes: DICE work. Separate working group. Decided to abandon. Made changes to DTLS which some people did not want. Implemented in RIOT. Need 6kB of memory. Works on every MCU device. @Michael: 6kB to get to end of the process. How much persist. Results are pretty good. @Hannes: If DH you need bignum. You need RAM otherwise stuff will not end. 10kB RAM for reasonable protocol run. Exchange only happens once. Providing guidelines for minimal server conf. Is there interest. @Carsten: When I would use this. I open a channel to server. I do a get for key. @Hannes: We have work on group communication. ACE working group commn. You went for wrong thing. Refers to a document. Minimal IKE doesn't do you favor. ACE add PK. @Brian: Other draft is still be formed. There is one implementation. Not for IoT implementation. @tero: Old implementaiton. Msec was shut down.There is AD support. Can go to IPsecme.