Clone wiki

meetings / 160902_webex

DRAFT

Minutes, 02 September 2016 interim, 6TiSCH WG

Note: timestamps in PDT.

Connection details

Taking notes (using Etherpad)

  1. Xavier Vilajosana
  2. Pascal Thubert

Present (alphabetically by first name)

  1. Diego Dujovne
  2. Francesca Palombini
  3. Georgios Papadopoulos
  4. Malisa Vucinic
  5. Michael Richardson
  6. Nicola Accettura
  7. Pascal Thubert
  8. Qin Wang
  9. Rashid Sangi
  10. S.V.R. Anand
  11. Tengfei Chang
  12. Thomas Watteyne
  13. Xavier Vilajosana

Action Items

  • Thomas to propose to the ML the mechanism in order to allocate TX/RX cells instead of only TX as it is now

Agenda

  • Administrivia [2min]
    • Approval agenda
    • Approval minutes last call
  • IETF update [Chairs] [5min]
  • update 6top [Thomas] [20min]
  • update security [Michael, chairs] [10min]
    • introduce new DT / New angle
    • Potential for both minimal/PSK (based on Art)
    • and a more ANIMA-like solutions (DT task)
  • Next steps for SF0 [Pascal] [20min]
    • Target: Experimental?
    • Cell allocation Scheme
  • AOB [1min]

Minutes

  • 7:05 Meeting starts
    • recording started
    • first meeting after IETF 96 in berlin
  • Administrivia [2min]

    • Approval agenda

      No issues raised. Agenda approved.

    • Approval minutes last call

      No issues raised. Minutes last call approved.

  • IETF update

    • minimal draft for many days in the IESG review queue
    • There are still reviewers that have not looked to the latest version
    • we expect reviews by 2 weeks from now.
    • milestones for 6tisch have been reviewed
      • 6top sublayer -> 6P
    • we need to agree the date to do the updates according to the progress of the draft
  • Update on 6top (Xavi)

    • changes in 6top since ietf meeting
    • Multiple changes made before summer break presented by Xavi at IETF
    • Thomas introduces further proposals. We believe that after this proposals the draft will be ready.
    • STATUS command
      • 6P status command replaces the COUNT cmd
      • Response contains 2 counters (cells from A to B and B to A -- A is the node that requests)
    • 6Top LIST it has a pagination problem. It did not work if we listed more than the cells that fit in a page.
    • in the new version of the command we can ask for a number of cells starting from an offset.
    • the answer is a list of cells.
    • Xavi points that Tero proposed an additional improvement; if we ask for too many cells they may not fit because of IEs that may need to be placed. So the number of cells should be a max number not the total number, best effort for the responder to put as many as it can.
    • Thomas asks clarification; Xavi adds that the size of headers depends on the IEs, so no guarantee that we can always place all the cells
    • Thomas indicates that the draft already has a meaning for RC 1, no more cells available; maybe need a new RC.
    • Thomas: Complicated one: consistency probme found at plugtest. If one node is not aware that it has a cell availabel, it will never be used and the error will not be detected. Solution is to have a generation number to figure the latest change. Since there is no concurrent transaction between 2 peers, a counter can detect mismatch. 2 bits is for lollipop behavior to detect a reset. At every packet neighbors check consistency, action on inconsistency s to be taken by the SF (clear is recomended)
    • Generation numbers are used to check the consistency between the schedule of two nodes.
    • the counter has 2 bits. this is enough as there are no concurrent transactions
    • the lollipop counter starts by 0 upon clear or start, then it goes from 1 to 2 or 2 to 1 but never reaches 0 unless there is a clear.
  • Update on security

    • when we issue a clear cmd, both nodes clear their schedules.
  • Proposals by Thomas:

    • enable the allocation of cells from the receiver side. This enables a parent to defragment or build a better schedule.
      • propose a TX/RX flag in the cell add command
      • Action: Thomas to propose to the ML the mechanism in order to allocate TX/RX cells instead of only TX as it is now.
      • Pascal: It is possible to indicate the latency in a child node so a parent may decide to reshuffle. We need a "Move" command
    • On a failed 6P ADD the response may add a suggestion for the cells to allocate.
    • Nicola: another option could be that the transmitter proposes all the available cells in the transmitting side.
    • the representation of the cells is ope to the SF so the format can be defined by the SF. The 6P only proposes a format but can be overriden.
  • update on security (Michael Richardson)

    • Need a simple PSK solution
    • we are now considering a variation
    • Maria-Rita:
      • shows a diagram with different nodes in a network, Some nodes aim to join. One node is malicious. One is misplaced in the network.
    • One slide shows the flows of information from the joining nodes to the JCE
    • JCE can decide to spread the join load in the network, sequentially, etc..
    • Thomas: how the JCE nows that there are pending initiations?
    • Maria-Rita: pledge initiated connection. We allow join traffic, we have a specific period where we accept join traffic
    • Maria-Rita: we want to avoid to waste bw and energy.
      • Idea: give each pledge an interval between retries. They use the best effort space for join traffic.
  • Nextext steps for SF0

    • No time for this item today. We'll schedule it for the next call on September 23.
  • AOB [1min]

    • No issues raised.

Updated