Clone wiki

meetings / 131011_webex

Minutes Webex 11 October 2013, 6TiSCH WG

Note: timestamps in PDT.

Taking notes (using Etherpad)

  1. Xavi Vilajosana
  2. Pascal Thubert
  3. Thomas Watteyne

Present (alphabetically)

  1. Thomas Watteyne
  2. Pascal Thubert
  3. Diego Dujovne
  4. Geraldine Texier
  5. Guillaume Gaillard
  6. Maria Rita Palattella
  7. Nicola Accettura
  8. Pouria Zand
  9. Qin Wang
  10. Raghuram Sudhaakar
  11. Rouhollah Nabati
  12. Tom Phinney
  13. Xavi Vilajosana

Recording

Slides

Action items

  • draft authors to rename the draft filenames and references to drafts from 6tsch to 6tisch.
  • Thomas to email Qin and Yoshihiro to ask to ask for comments on the 6top/security drafts.
  • Pascal to discuss on ML whether it makes sense to have a timestamp with configurable granularit
  • Qin to follow up profile definition thread on the ML.

Agenda

  • Administrivia [3min]
    • Approval agenda
    • Approval minutes last call
    • IETF 88: final agenda today
  • 6TiSCH WG formation [10min]
    • 6TiSCH is a WG!
    • Practical changes
    • Next steps: publish drafts
    • tools: trac vs. bitbucket
  • update draft-sudhaakar-6tisch-coap [25min]
  • on-the-fly scheduling [20min]
    • update
    • discussion
  • AOB [2min]

Minutes

  • [08.05] Meeting starts
  • Administrivia
    • Approval agenda

      No issues raised. Agenda approved.

    • Approval minutes last call

      No issues raised. Minutes approved.

    • IETF 88: final agenda today
      • cut-off date for drafts is 21-10-2013 (all, -00 and not -00).
      • Final agenda should be published today (11-10-2013)
      • 6TiSCH meeting scheduled Tuesday Nov 5th 1420-1550 (THIS IS TENTATIVE!)
      • usual e-mail from Carsten with summary schedule, about all IoT related WGs sessions
  • 6TiSCH WG formation
    • 6TiSCH is a WG!
      • 6TiSCH approved as official WG!
      • Thanks to Ted Lemon, Adrian Farrel, Marc Blanchet, Zach Shelby and Benoit Claise.
    • Practical changes
    • Next steps: publish drafts
      • [Pascal] please publish basic drafts first as others can update references.
      • Wait until terminology draft is published. It takes some time so the cross-reference utility to catch up with new submissions.
      • [Thomas] No difference on cut-off date for -00 and non-00 drafts.

        Action item: Thomas to email Qin and Yoshihiro to ask to ask for comments on the 6top/security drafts.

    • tools: trac vs. bitbucket
      • Trac environment in IETF home page
      • We have redundant tools
    • [Pascal] 6lo charter got approved today. Great news!
  • update draft-sudhaakar-6tisch-coap
    • https://bitbucket.org/6tisch/draft-sudhaakar-6tisch-coap
    • summary discussion
      • Raghuram gives an update
      • content of the draft:
        • what type of methodology/protocol to express rules to define extensible resources? How users are able to define specific events (ad-hoc events)?
        • Regular expressions: do not restrict to simple expressions, ABNF is considered.
        • Complexity and capability of HW devices to manage ABNF. If not supported, profiles to support or not extensible resources.
        • A collection of user-defined URIs is a profile. We can have a predefined "basic" profile.
        • New section. Generic Data Model. We do not want people to think that CoAP is the only mechanism to deal with interaction.
          • the section will provide a generic data model so other "transport" protocols can be used.
          • the section will be empty for the Vancouver meeting
          • the section will be expressed in the Yang language.
    • NVP vs. CBOR
      • Carsten pointed out that CBOR is a useful tool as similar parsing is required in NVP.
      • Xavi looked at Carsten's preliminary C implementation of CBOR to assess its complexity and see if it could run on typical LLN micro-controllers
      • code is mainly xors and shifting so should be OK. No complex structure that could be an issue.
      • Only thing is 64bit variable which have overhead on 8/16-bit micro-controllers.
      • [Pascal] discussion about representation of ASN in CBOR. One way to use a string format (5-byte string).
      • [Pascal] ASN is 5 bytes. As timestamp we can use some of the most/less significant bits if this is a problem. Depending on granularity or precision requirements in timestamp.

        Action item: Pascal to discuss on ML whether it makes sense to have a timestamp with configurable granularity based on ASN. Differentiate the ASN from the timestamp?

      • If it is an string we have to manipulate it. This is hard to manipulate.
      • [Thomas] A string is just a succession of bytes. Could be manipulated like a buffer. Dealing with endianness is very common.
    • Table relationships:
      • [Pascal] Models relations of tables
      • Help us optimize the organization of the information.
      • 6TiSCH resources: How do we represent that information in the draft? tables? YANG?
      • [Thomas] is this more a 6top data representation and not a CoAP representation.
      • [Qin] tables come from 6top. More relation to 6top. It is a representation of the 6top MIB.
      • [Qin] 6top MIB needs to be reviewed how to add bundle/track etc. and other features on the MIB.
      • [Raghuram] this has implications on the URI naming: if there are changes to the MIB, CoAP URIs need to be changed.
      • [Pouria] 6TiSCH resources section in draft: what should be shown on that section?
      • [Raghuram] informational resources: variables that can be GET. Management resource: variables that can be used to configure and manage.
      • [Raghuram] The PCE is only allowed to read from Informational resources (read only).
      • [Thomas] this goes beyond the PCE talking to CoAP, also will apply to On-the-Fly scheduling for example. Use the same mechanism in any case.
      • [Raghuram] Agreed.
    • customize functions in user profile

      Action item: Qin to follow up profile definition thread on the ML.

  • on-the-fly scheduling
    • [Diego] Update about ML discussions
      • challenges:
        • find input parameters for the algorithm.
        • adatpive algorithm itself.
        • allocation policy (collision avoidance)
      • Parameter information to use: historical trends, last value, combination of parameters.
      • Algorithm: open to implementations.
      • Allocation policy:
        • pre-allocation: use a bundle of cells (some being used or not)
        • post-allocation: on response if resource needed
        • hybrid: based on information from historical and actuate w.r.t the last value.
      • How to have parameters open so different algorithms can be implemented.
      • How to trigger algorithms? What nodes are candidates to start allocation dynamically?
      • How to avoid collisions?
      • Map of cells? global picture? 6top? management module?
      • pre-allocation: who keeps the map of allocated slots if there is no PCE Limit of preallocated cells in the bundle vs. side of the schedule?
      • How many pre-allocated cells should be in a bundle?
    • [Thomas] Quick comment about scope of this work w.r.t. 6TiSCH. There will be several algorithms of how to find out when to add/remove cells in a bundle. Assuming we have an algorithm: what information do neighbor need to exchange? Is what is now in 6top enough? do we need more mechanisms? Important part is what goes over the air. Make sure that 6top mechanism are sufficient to support that.
    • [Pascal] Agreed. The work on OTF is very important, also to understand the ways 6top will be used. This will trigger requirements for 6top.
    • [Nicola] Cell reservation is distributed.
      • TASA- Centralized.
      • DeTAS-decentralized: Traffic aware.
      • [Maria Rita] TASA is not very different from OTF. TASA builds the schedule in a centralized way but using On-The-Fly information.
      • exchange of packets between neighbor nodes -
      • OTF proposal: take traffic by each node and children nodes into account.
      • Child nodes sent schedule to its parents. After receiving its child's schedule, parent update schedule and sends it to its children.
      • Dynamic adaptation: a flag in the data frame to be sent in the first TX cell of the bundle.
      • Child/parent schedule. They can be mapped in the Schedule IE if a new TLV element is added to the IE.
      • [Thomas] you need a Hard Cell reservation from neighbor to neighbor.
    • discussion
      • to continue on the ML
  • AOB

    No other business raised.

  • [09.07] Meeting ends

Updated