Minutes IETF97: roll

Meeting Minutes Routing Over Low power and Lossy networks (roll) WG
Title Minutes IETF97: roll
State Active
Other versions plain text
Last updated 2016-11-21

Meeting Minutes


11:10 - 12:10  Wednesday Morning session II

Status of the working group ------------------------------------ 11:10  - 11:18
 (8mins) --- Peter/Ines Use of rpl info draft - 
draft-ietf-roll-useofrplinfo-09 ------- 11:18  - 11:26  (8 min.) -- Michael
Richardson AODV-RPL draft - draft-satish-roll-aodv-rpl-02 -----------------
11:26  - 11:41  (15  min.)- Charles Perkins DAO projection:
draft-thubert-roll-dao-projection-03 ----------- 11:41  - 11:55  (14 min.)--
Pascal Thubert Mpl Forwarder select - draft-vanderstok-roll-mpl-forw-select-02
-11:55  - 12:08  (13 min.)-- Peter van der Stok Q&A 
----------------------------------------------------------- 12:08  - 12:10  (2
min.) -- Peter/Ines

Taking notes

* Dominique Barthel


* [11:12] meeting starts
    * agenda bashing, blue sheets, etc [Chair]

* [11:13] status of the working group (Chairs)
    * Peter: very few comments on WG LC. Please read and comment the drafts on
    the mailing lsit * Inès: shows status of drafts, WG and individual * Peter:
    several individual drafts will be presented/discussed dhere. We'll ask for
    WG adoption so that they can move forward * Inès : charter update. Please
    read (see mailing) and comment. * Peter: if nothing heard, will push
    changes to IESG.

* [11:16] Use of rpl info draft -  draft-ietf-roll-useofrplinfo-09 -- Michael
    * discussion at 6man, RPI can stay across the edge of LLN, will not be
    processed. * read and comment * Peter: who has read the draft? xx hands
    [xxx] can you explain why decision RPI could be left in place? [Michael]
    originally, packet with RPI was to be dropped. Intended to contain trafic.
    Now, because we are not required to remove the header, there are situations
    where we can avoid IP-in-IP encapsulation. [Pascal] new 6loRH encoding
    might change things a little bit. Still discussions at 6man. Update of 2460
    will not change things drmatically. Also, discussion on when to insert
    headers is somewjhat different from when to remove headers. RH3, when
    compressed with 6loRH, .... RPI header come first, could be cut-off
    [Michael] could happen. Doc says do RPI, not IP-in-IP, for upward path.
    Decision to remove RPI is at the root, which is more flexible/capable/...
    [Michael] still breaks authentification header. But confident we can send
    them out to the Internet and not break the architecture. [Pascal] to the
    fact that it breaks AH, .... [Micahel] agree that rules for remove=ing
    header are weak. We should write a separate document. Decidable by the 6LBR
    independantly of any other decision [Pascal] could be written in here.
    recommendation would be : 6loRH, strip it out. [Michael] feel this would
    drag this document into 6man discussion. Let's write this is separate
    document and wait for 2460bis to settle. [xxx] don't know what's going to
    happen at 6man. [Michael] think we can easily write it as a separate
    document. No effect on the little devices. Peter: please read the doc and
    comment on the mailing list.

* [11:30] AODV-RPL draft - draft-satish-roll-aodv-rpl-02 -- Charles Perkins
    * extension to P2P-RPL using AODV
    * describes recent changes to draft
        * pairing on Route Request DODAG instance and Route Reply DODAG
        instnace by odd/even instance IDs * this saves Destination IP address
        in RREP message
    * implementation on Contiki/Cooja
    * expect demo at next IETF
    [Ravishankar] what model for Contiki/Cooja simulation?
    * chairs: who read the drat? 5 hands. Who wants to review? 2

* [11:39] DAO projection: draft-thubert-roll-dao-projection-03 -- Pascal Thubert
    * we can turn RPL into centrally controlled algorithm, e.g. for factory
    networks. * establish special paths with "garanteed" capabilities. E.g. for
    6TiSCH. * PCE tells root with route to install. * another benefit is to
    compress source routed path. Turn strict source routing into loose source
    routing. Since done centrally, careful decisions. [Ravishankar] cost of
    computation? [Pascal] in Non Storing Mode, topology is built. Unchanged.
    Just compressing the expression of the path in Source routing. * version 2
    of this draft introduced Storing Mode variation. But additional protocol
    elements needed. In this draft or another draft? * allow to build
    transversal pathes. * new on -03: not much, some cleanup [Ravishankar]
    discussion years ago on mailing list about mixed mode. Does this draft
    allow it? [Pascal] in a controlled way. Risk of storing mode is when nodes
    run out of memory. There has been discussions on refusing DAOs. Hard to
    make the distributed approach to this. Centralized approach works. PCE
    knows amount of state, lifetime, etc. Pascal: many more slides provided to
    illustrate how this works Peter: who has read? 6 who is willing to review?
    4 who thinks the WG should work on this? 6 <<>>>

* [11:52] Mpl Forwarder select - draft-vanderstok-roll-mpl-forw-select-02 --
Peter van der Stok
    * problem is selecting a few forwarder of nulticast traffic, for
    reliability. * Peter explains algorithm, see slides * goes through
    examples, using diagrams * Peter: interest in the WG in this work? [Charlie
    Perkins] this is great idea. See also work at MANET. Connected Dominating
    Set. [Charlie] your picture show regular arrays. Might be misleading.
    [Peter] agreed. Much more work to be conducted. [Charlie] is this meant for
    stably establisehd networks? [Michael] please explain how far the nodes can
    hear one another on theses disgrams. [Peter] on slide 11, radius of 3
    lattice edges. [Michael] does this live on top of MPL? [Peter] on top of
    Trickle. [Michael] at BIER, they might be interested in this. [Peter]
    discussed it with Carsten. [Pascal] interested in this. Variation on
    MultiPoint Relays of OLSR. However, on variable radio conditions and
    bsttery limited devices, you might drain device batteries. How about a
    rotation mechanism so that ndoes take turns? [Peter] excellent idea. Was
    thinking of moving forwarders around to get further local optiisation Inès:
    who has read? 2

* [12:07] Q&A  -- Peter/Ines
    * any question? on drafts? on charter?
    * no question
* [12:07] meeting closed