Clone wiki

meetings / 160122_webex

Minutes, 22 January 2016 interim, 6TiSCH WG

Note: timestamps in PST.

Connection details

Taking notes (using Etherpad)

  1. Thomas Watteyne
  2. Pascal Thubert
  3. Diego Dujovne
  4. Keoma Brun-Laguna
  5. Michael Richardson
  6. Xavi Vilajosana
  7. Simon Duquennoy

Present (alphabetically)

  1. Thomas Watteyne
  2. Pascal Thubert
  3. C-Y Lee
  4. Diego Dujovne
  5. Jonathan Munoz
  6. Keoma Brun
  7. Michael Richardson
  8. Pascal Thubert
  9. Patrick Wetterwald
  10. Qin Wang
  11. S.V.R. Anand
  12. Sedat Gormus
  13. Shahid Raza
  14. Simon Duquennoy
  15. Simon Duquennoy
  16. Sven Akkermans
  17. Tengfei Chang
  18. Xavi Vilajosana
  19. Zhuo Chen

Action Items

  1. QIN: will start a discussion on the ML on what we do on information and yang data model

Agenda

  • Administrivia [2min]
    • Approval agenda
    • Approval minutes last call
    • rechartering news
  • 6LoRH [20min]
    • draft(s) status
    • ML issues, proposed formats
  • 6top-sublayer [10min]
    • ML Discussion on recommended formats
  • Minimal Draft [15min]
    • impacts on main text from intro changes
    • Suresh's issues
    • Next steps
  • PlugTest [10min]
    • Walk-through (delta) tests
    • related question about the 6lorh and 6top
  • AOB [1min]

Minutes

  • [07.04] Meeting starts

  • [07.07] Administrivia [2min]

  • Approval agenda

    • no issues raised, agenda approved
    • Approval minutes last call
    • no issues raised, minutes approved
    • rechartering news

      • mcr was asked opinion by IESG about 6tisch recharter, wrt ROLL and mcr confirmed it was good.
      • Xavi: we need some information about the layer
      • Pascal: let's not confuse terms
      • Xavi:
      • Thomas: Yang model is extensive, should we start from Draft reduce to the security ?
      • Pascal: question from Benoit is confusing a little bit. Where is the information model ?
      • Thomas: data model = interface draft whereas information model = sublayer draft
      • Qin: suggest we discuss in ML
      • Pascal: can you start thread
      • Qin: OK (ACTION)
      • Thomas: Can we avoid this discussion again ?
  • [07.15] 6LoRH (Pascal) [20min]

    • draft(s) status
      • Thomas:
    • Pascal: 6LoRH draft adopted
    • topic: how do we go certain things, answers tech questions from Tengfei and Simon

      • example 1: case where we have a packet from the DODAG toward the Root (ICMP and UDP)
      • In this case we don't need IP-in-IP ???
      • Where is the RPI ?
      • slide 9: ICMP example
      • slide 10: UDP example
      • 6LoRH goes BEFORE the IPHC, this is an important addition of the draft
      • node which wants to add a 6LoRH will NOT have to change the IPHC packet
      • things are reversed: IPHC and IP-in-IP goes at the END of the packet
      • as packet progresses, the addresses in 6LoRH are being removed. Different than IPv6. In
      • in 6LoRH:
      • routing head is "consumed": size of 6loRH is reduced as the packet travels from root into the LLN
      • the next hop is the first address in the 6LoRH, NOT the IPv6 destination address (final destination in inner IP-in-IP header)
      • Thomas: Do we have to carry the full IPv6 address in all RHs?
      • Pascal: "normal" case in 6TiSCH:
      • only 2-byte addresses, remaining 112 bits inferred from the source
      • you know what the source of the packet is
      • Simon: about slide 11 removing addresses as you go is good. But how interoperable with RFC6554? all addresses are in RFC6554? Would expect from 6LoRH to compress existing 6554 headers.
      • Pascal: you can reconstruct an RH3 from 6LoRH, but it will not be the same one as if you
      • Michael: important for efficiency. It's possible to deploy nodes with 6LoRH but not turn it on (manageable)
      • why does it matter?
      • Simon: in the routing header, the addresses are not consumed
      • Pascal: 6man might complain. problem with authentication.
      • Simon: big problem?
      • Pascal: we have L2 security, we can trust that routing header will not be compromised.
      • Simon: big sec threat. 6LoRH should not impact 6LoWPAN
      • Pascal: do you want to keep the data in the RH? -technical decision
      • Simon: AH (authentication header). In IPv6, you have to replace the destination address.
      • Pascal: Any RPL packet could be attacked the same way. But it is not because we have L2 security. If we wanted to keep the addresses we could add an index in 6LoRH to ... WG to decide
      • Simon: why couldn't we remove addresses from 6554?
      • Pascal: probably for passing 6man faster.
      • Thomas: the idea of 6554 is to do a version for 6LoWPAN. When we implemented the last plugtest, we realized that we could ping 3 hops away
      • Even if we add an index, we will not pass 6MAN if we have an option not to use it ??
      • Xavi: 6554 doesn't say anything about what happens
      • the receiver doesn't ...
      • Thomas: we can specify how the AH is calculated
      • Pascal discusses different use cases with the types of the addresses.
      • the new draft explain the recursive process of popping the first address in RH3-6LoRH, asking text review

      • If the packet is fragmented, same fragmentation rule as IPHC

      • discussion about where IP-in-IP header goes. Slide 14 about proposed order.
      • when decoding 6LoRH headers,
      • Thomas: Why multiple RH3?
      • Pascal: e.g. from one PAN to the next
      • Xavi: clarifying: fragmentation header right after MAC?
      • Pascal: yes
      • Pascal: without order, you cannot do IP-in-IP-in-IP
      • Pascal: discussion is about the order of the headers, proposal:
        • 1st reason: starting with RH is simpler as it is consumed
        • 2nd reason: need a clear separator
      • Pascal: proposal: reverse all the headers (keep logic as with IPHC and 6LoRH) IPHC and 6LoRH are the separators
      • Pascal: RH3, then RPI, then 6LoRH.
      • Simon feels it adds complexity
      • the draft is redesigned to work on the compressed form
      • if you don't want that you have to rewrite everything
      • Thomas: Do we remove addresses of the routing header ?
        • What are the arguments to put the header before ?
      • Pascal: the main idea was to make easy to manipulate the header
        • When its compressed its not meant to be read. DO people read in the zipped file?
  • [08.09] PlugTest [10min]

  • Thomas: motes sent to the people
  • dissector available
  • Golden version available
  • Walk-through (delta) tests

  • related question about the 6lorh and 6top

  • poipoi
  • [07.??] AOB [1min]
  • poipoi
  • [08.??] Meeting ends

Updated