MBoned WG IETF 79 (Beijing)
Wed, Nov 10, 2010
13:00-1500, Emerald room
Audio archive of this meeting: http://www.ietf.org/audio/ietf79/ietf79-ch1-wed-noon.mp3

Marshall chairing this time

Status update

ipv4-uni-based-mcast is now RFC 6034!

mtrace draft finished wg last call
  Hitoshi:
  reviewed by AD, he got many editorial comments
  some text not clear
  Ron wanted some rate limitation. Hitoshi will
  revise soon

mboned auto multicast
  0 checksum doc accepted in 6man. this is only
  thing amt is waiting for

[**Note from Lenny- there are actually a few other items 
brought up in Anaheim (teardown, sourcing) that need to be resolved with AMT]

Marshall asking people to review and comment on
  the AAA documents.

[some other drafts also briefly mentioned]

Explicit tracking update
  Marshall explaining about multicast protocol issues
  like IGMPv3 and how to find a home. mboned has been
  doing protocol work even if an ops group. Mentions
  that pim wg may take on protocol work and has agreed
  to do igmp work like explicit tracking.


draft-cociglio-mboned-multicast-pm-00
  A method for multicast performance monitoring
  Stig, why does it have to be self contained?
  Marshall, why not for instance use RTP if you happen
    to have an RTP stream, why not use the information.
  Someone asked some question on jabber

  Marshall, where do you put the timestamp? Are you
  doing just one bit?

[ATB] The bit (f.i. in DSCP field) is just used to mark the traffic to be monitored and determine when to take the timestamp.
Timestamp itself
is not stored in the packet header but sent to an external system which performs the actual measurement.

  Marshall and user online comments on timestamps and
    synced clocks for delay
  Author: they have done it for packet loss. they are
    still looking into time stamps and checking jitter
    etc.
  Ilya Valashkin (sp?) online says trials needed for
    timestamps, it can be complicated.
  Toerless, says he wonders what is the best wg. He
    mentions IPPM. It has already been there though?
  Stig, we can in this wg say what measurements we need
    but maybe not how
  Iliya mentions MPOL?

[ATB] PMOL work is about performance metrics for other layers, IP performance metrics are discussed in IPPM WG.

  Marshall, this group could do requirements
  Stig, question about reordering and marking
  Author: not an issue
  Author: figured that IPPM best place...
  Vero: there is interest in IPPM, but chair decided
  not sufficient to take it on.




Requiremnts for PM, Vero
  Marshall, this can be useful in many areas. Some
    things can be very general.
  9 think requirements for PM should be done in mboned
  0 against

Dual Mtrace2, Hitoshi
   Toerless, what is special, why not two separate
     traces...
   Stig, trying to respond, for backup paths it is
     not visible to end user, so network must take
     care of it

Tuning IGMP and MLD
   Stig asked people to review both the multimob drafts
     on this topic and post comments to the multimob list.
   Stig said that spec allows unicast queries, but not
     sure if implemented. He asked if anyone has
     implemented this.
   Marshall asks if it is a problem reducing the robustness
     value. Marshall explains how it may be a problem
     with fast leave and low robustness variable.
   [[Hitoshi answered it could be possibly safe in the
     good-condition link with shorter general query interval.
     The draft mentions it.]]
   Toerless: unicast queries are dangerous, they may be
     sent remotely
   Stig: Yes, dangerous and I think [[IGMPv3/MLDv2]] spec say they
     must be accepted. Maybe spec should be updated.
   [[Hitoshi: Can be written in the appendix in my tuning draft.]]

Channel Reflector
   
Discussion about mboned-session-announcement-req

  Marshall says a new last call may be needed.
  Marshall trying to ask the room about support.
  Think one person in support
  Also Stig supported requirements but don't remember
    document.
  Lenny on Jabber says something about not enough support.

[**Note from Lenny- in April, SAP requirements draft went 
through WGLC.  Only 4 people responded- 1 in favor and 3 opposed to advancing the draft.  This was determined to be insufficient support for advanding the draft.]

  Other author (Vincent Roca) says some people are still using
    SAP and good to document [[SAP limitation because some
    applications, e.g. discussed in FECFRAME WG, assume to still
    use SAP. At least, the document clarifying the issues and its
    limitation must be published.]]
  Toerless, one thing he didn't see is that it is light
    weight, not requiring any infrastructure etc...
  Hitoshi says [[the goal and aim of this draft is not for new
    proposal or protocol. CR is one of the concrete implementation
    fulfilling the requirement.]]
  Toerless, asks whether requirements are needed
  Hitoshi, thinks should collect requirements because might
    want to develop new solutions, e.g. CR
  Marshall, there has to be more enthusiasm for it to be revived.
  Toerless, some may need service discovery in the network.
  [[Hitoshi says he hasn't expected to do draft on CR because the most
    important thing is to clarify the requirement. But if some needs,
    he would draft the solution such as CR.]]
  Stig, SAP can be useful in certain environments