1. 6tisch Avatar 6tisch
  2. admin
  3. meetings

Wiki

Clone wiki

meetings / 140408_webex_6top

View History

Minutes Webex 8 April, 6top call 6TiSCH WG

Note: timestamps in PDT.

Taking notes (using Etherpad)

  1. Xavi Vilajosana
  2. Thomas Watteyne

Attendance (alphabetically)

  1. Pat Kinney
  2. Qin Wang
  3. Raghuram Sudhaakar
  4. Thomas Watteyne
  5. Xavi Vilajosana

Agenda

  • exercise on implementing 6top negotiation with YANG and CoAP
  • using CoAP to implement 6top management interface
    • How to express node address? use uri-host or not? IP address or MAC address?
    • Need observer function? use uri-port?
    • How to fragment data to fit IE payload size? use CoAP block?
  • 6top-to-6top Access Control model.
  • How to express 6top MIB and IEEE802.15.4 PIB in YANG? (comments from Juergen)

Slides

Action items

  • Qin to create a ticket before editing 6top-interface draft.
  • Raghuram to work further on how to proceed w.r.t fragmentation.
  • Qin to start thread on the ML about 6top-to-6top Access Control model.
  • Thomas to start thread on the ML about how to express 6top MIB and IEEE802.15.4 PIB in YANG.

Minutes

  • _[20:05]_meeting starts
  • [Thomas] note well, minutes, attendance, etc.
  • [20:07] Qin presents the slides. Implementation exercise of the soft cell reservation process.
  • [Qin] Presents the steps to negotiate for cells between 2 nodes.
    • BWIE and ScheduleIE are used. BwIE to indicate the number of cells required; ScheduleIE to indicate which cells are candidates and eventually selected
    • Soft cell reservation response includes the number of reserved cells and the list of cells. Uses same IEs.
  • [Qin] Question is how to specify this operations within the scope of CoAP.
  • [Raghuram] Why not have the response directly in POST response?
  • [Xavi] Maybe to stick closely to CoAP spec?
  • [Qin] let's go through other options
  • [Qin] 3 possible schemes:
    • 2 resources defined: one for reservation request, another for response.
      • POST, then GET (2 operations)
    • 1 resource POST and response
      • [Qin] problem is how to separate parameters and response in the YANG model
    • 1 resource GET. Query by parameters. Response is an object representing the result.
  • [Thomas] why a payload associated to a YANG model? Consistency of the messages across different protocols. The YANG model does not say what the payload is, it just describes the MIB.
  • [Raghuram] The question if we need to maintain consistency of this model across different protocols. If we use RSVP we need to be able to use the same semantics.
  • [Xavi] do you think the format is important, or the content? Would YANG not be translatable to different protocols?
  • [Raghuram] might be disadvantageous to be too specific in YANG
  • [20:21] [Raghuram] Having 2 packets instead of 4 is an advantage from 2nd approach.
  • [Raghuram] CoAP response can be asynchronous so if computation of the schedule takes time still works.
  • [Thomas] CoAP has a capability to have requests with separate responses. Both piggybacked responses or delayed responses are possible.
  • [Raghuram] proposes to keep the GET NegotiationResult as an optional operation to be able to query latest result and in case the response is time out.
  • [Thomas] Representation of what we want to do using YANG model as RPC.

    https://tools.ietf.org/html/rfc6020#section-4.2.9

  • [Thomas] The CoAP payload then will be represented by this YANG model specification.
  • [Xavi] Express the reservation process can be expressed using RPCs and then this will be supported by any transport mechanism. e.g rsvp, coap, xmpp, etc..
  • [Thomas] RPC is meant to express this kind of interactions.
  • [Raghuram] section 7.1.1 summarizes sub-statements.
  • [Thomas] Optional fields?
  • [Qin] Need to be check. Some interactions do not need schedule IE, so how to make that optional.
  • [Xavi] fields cannot be marked as optional, only mandatory (RFC65020, section 7.9.4).
  • [Thomas] We need to be very clean when editing the draft. As it is a WG document now, we need to explain everything to the ML. Create a ticket is a good idea as well.

    Action item: Qin to create a ticket before editing 6top-interface draft.

  • _[20:44]* using CoAP to implement 6top management interface
  • How to express node address? use uri-host or not? IP address or MAC address?
    • [Xavi] Use MAC layer address as it is a L2.5 thing.
    • [Raghuram] +1
    • [Thomas] Remove it completely as MAC layer addresses are at L2 so we can use them.
    • [Raghuram] It is a good idea to keep mechanism identical to L7 CoAP, but still addresses can be inferred from l2 information as in the case of 6lowpan.
    • [Thomas] Come back to that issue when more progress in ACE.
  • [20:50] Need observer function? use uri-port?
    • [Thomas] What is the relation of uri-port and observer function.
    • [Qin] Port might be used to indicate which observing server is sending the observe.
    • [Thomas] That's not right, that's what the token is for.
    • [Qin] We need observe in PCE case: if some data changes it will trigger some recalculation on the PCE. Analogously, the 6top-to-6top case might require to trigger modifications on the neighbors, this might be dependent of the algorithm used.
    • [Thomas] Observe is optional, so if it is not present, there should be a mechanism to know if the mechanism is running on a particular node. To be checked.
  • How to fragment data to fit IE payload size? use CoAP block?
    • [Raghuram] CoAP Block, we need to come to a decision.
    • [Raghuram] Limit the payload length. Candidate list needs to be restricted. We need to determine how often this case will happen.
    • [Thomas] We need to make sure that everything will work taking into account that packets are 127B long.
    • [Thomas] As we need to reassemble, we need to have buffers to be able to reassemble at a node, this requires RAM, We need to take that into account if we want to be able to run in any very constrained device.

      Action item: Raghuram to work further on how to proceed w.r.t fragmentation.

    • [Thomas] CoAP header and IE go inside the mac layer payload. UDP headers are not considered.
  • [Thomas] Out of time, continue on ML.

    Action item: Qin to start thread on the ML about 6top-to-6top Access Control model.

  • How to express 6top MIB and IEEE802.15.4 PIB in YANG?

    Action item: Thomas to start thread on the ML about how to express 6top MIB and IEEE802.15.4 PIB in YANG.

  • [09:07] meeting ends.

Updated