LWIg @IETF 96 12:20-13:20 Friday Afternoon session I Room: Schoeneberg Berlin Intercontinental * Minute takers: Rahul * Jabber Scribe: Renzo Navas 1. Agenda bashing 5 min Zhen: presenting the agenda 2. TCP over Constrained Nodes Carles Gomez Montenegro https://tools.ietf.org/html/draft-gomez-core-tcp-constrained-node 20 min slides: https://www.ietf.org/proceedings/96/slides/slides-96-lwig-1.pdf Carles: TCP over constrained node networks TCP will be used in IoT networks ... TCP has not recevd much attention... Goal is to offer simple TCP measures for IoT networks Explains CNN characteristics (RFC 7228) Recommendations about using TCP over CNN... possible recommendations in the form of TCP MSS. Recommended use of one window size Performance evaluation of CoCoA mentioned , recommending use of CoCoA recommendations on the keep-alive mechanism .. Consider use of TCP fast open .. Need some detailed anylsis ... Cookie is included in the pkt increasing the msg size ... needs detailed analysis. ECN can be used to trigger congestion control mechanism earlier ... Options in TCP which would not impact in such cases (LLN cases)... SACK may be a relevant option. For future revisions, major scenario to consider is constrained node talking to non-constrained node. It will be great to know implementation experiences .. [Carsten]: its good thing to have doc like this .. doc could be little be more explicit about asymmetric implementations ... anything a less constrained implemetnation can do to help non-constrained impleemteations. we often forget, that another IoT protocol used is XMPP .. Oleg: draft about tcp compression... have the authors checked out this draft ?? [Carles]: had seen the draft, but several years old .. it would be nice to have tcp compression Oleg: TCP implementation as part of RIOT in progress kerry: Nodes have to be able to assemble mtu of 1500 .. pls dont limit to certain implementation ... TCP implementation already in contiki Michael: as TCPM WG co-chair. Interesting doc on how to implement. fail to see any value in changing the tcp spec in this wg. any change to tcp has to be taken up in appropriate in TCPM wg Suresh: Slide 12 doesnt belong here ... If WG decides to adopt the doc .. then the review has to be done from tcp group .. . Further discussion on ML... [zhen] agree with Michael on TCPM wg synchonization 3. Implementation experiences of public-key cryptography on 8-bit micro-controllers Mohit Sethi https://tools.ietf.org/html/draft-aks-lwig-crypto-sensors-01 15 min small devices of 8b uprocessing, 2-5kb ram ... performance undesireable with RSA .. signing with private key was pretty slow.. ECDSA implementation considered were for 8b uprocessors.. In <3KB RAM, you can do signing operation is less than a second.. shows numbers for signing and verification for 8b uprocessor... shows an eg app scenario .. details about the topology and configuration .. Can get quite minimal implementation with some efforts.. Cost is more to turn on/off the radio than the crypto operations!! Tradeoffs discussions ... misconception that sym crypto doesnt scale .. for low latency consider using object security .. ll2 security has problems .. because they use group keys usually .. Need to ensure freshness, to avoid replay attacks ... some guidelines on the same in the draft. Authors believe its fairly stable .. [Juan]: what kind of radio interfaces were used ? [mohit]: mostly ethernet based interfaces ... no radio.. ** About a dozen people have read the draft ... [raghavendra]: found useful for the thesis ** More than > 20 people think that the doc can be adopted ... ** No one against the adoption. 4. Minimal ESP Tobias Guggemos (Started 13h00) https://tools.ietf.org/html/draft-mglt-lwig-minimal-esp 10 min explains context to use IPSEC based security... Problems with DTLS, app layer security ... highlights privacy issues... with ipsec on the gateway, the ip addresses of end devices not sent.. IPSEc also capable of multicast security and any other transport layer protocol .. explains flexibility of IPSEC, any key exchange protocol can be used .. Goal of minimal esp is provide guidelines to implemnent IPsec for IOT networks... explains ESP packet format. .. explains use of the fields .. SPI, Seq Num, padding recommendations ... IPSec is a valuable option for IOT.. [suresh]: have we got any comments on the draft yes, Tero is involved .. [brian]: havent really cut any bytes ... no compression involved .. only implementation guidelines .. no compression included [suresh]: if the wg decides to adopt, suresh will help to get additional review done.. [zhen]: what's the difference of this draft with two other 6lo drafts. [daniel]: explains difference between diet ESP and this guideliness .. [zhen]: is any implemetation present? started implementation on contiki ... *** 6/7 people have read the draft.. Go on the list to check the consensus .. and hope to get reviews from security folks .. 5. Lwig terminology update discussion Carsten Bormann https://tools.ietf.org/html/RFC7228 10 min checks in room who knows rfc 7228 ? explains the need to update the rfc 7228 .. terminology needed to distinguish ucontrollers.. Two more classes to be used for distinguishing .. category based on crytpo supported, protection capabilities (secure storage, tamper-proof etc ) [Renzo]: On ACE we are facing problems related to clock capabilities (real time -wallclock-, relative time, no clock at all, ..) OAuth uses timestamp for token validations but some devices might not have. ... need clear defntion of clock capabilities, to clearlystate the problem, then define solutions. would like to assemble a group of people who can contribute ... [kerry]: extremely useful work ,, people think raspberry pi as constrained device .. not enough awareness on the power constraints ... [chistian]: note that there is work going on on tranport negotiation [ana miraburo]: (LPWAN) new scale of bytes transreception (talking about lpwan networks, very few payload/bandwith). TOn lpwan the thorughpout scale is one order of magnitude different from seconds to day [emmanuel]: support this work .. two phases in the doc to asses what is changed and then reduce the number of classes .. [suresh]: rfc is the right format for this ?? something which is evolving .. rfc 7228 is too young to be obsoleted.. wg should consider this point .. [carsten]: content needs consensus, so at least wg draft is a needed format, not wiki [zhen]: call for contribution .. [suresh]: can go for consensus without the draft