[Tsvwg] Draft Meeting minutes from Montreal

["James M. Polk" <[e-mail]>]

TSVWG

Here are the DRAFT meeting minutes from our Montreal IETF.  Please clarify 
where needed (sending comments to Lars too).

TSVWG Agenda for IETF 66 (Montreal)
Monday July 10th, 2006
0900-1130

Chair's                                                      9:00-9:15
       Agenda Bashing                                         (15 min)
    - we switched two presos in order
    - Bob's re-ecn-apps wasn't submitted on time, but the topic
      is still in the re-ecn-tcp ID
       NOTE WELL
       Document Status and Accomplishments
       New Charter
       New Milestones

Kwok - Aggregation of DiffServ Service Classes               9:15-9:25
   draft-ietf-tsvwg-diffserv-class-aggr-00                    (10 min)

- reviewed changes to this version
- all info should be in this doc, and not plan another follow-on doc
- wants reviews (David Black had already agreed)
- ADs stated

Randy and Michael - SCTPbis and SCTP Padding Chunks          9:25-9:40
   draft-ietf-tsvwg-2960bis-02                                (15 min)

- Randy discussed summary of changes
- discussed summaries/reviews made
- needs a final read
- may be some additional XML changes no one has caught yet
- asked for WGLC - to spur real reviews

   draft-ietf-tsvwg-sctp-padding-00

- summarized changes since last version
- asked for WGLC

Marushin - SCTP ASCONF Chunk Transmission Ext.               9:40-9:45
   draft-marushin-sctp-asconfext-01                           (5 min)

- reviewed what this ID is about
- this is an extension to SCTP ADD-IP
- reviewed a scenario in which this idea applies (regarding a IPv4 and
   IPv6 addressing mismatch on each end of a e2e flow, ext keeps both v4
   and v6 addresses on one side cached to make sure communications can
   remain up)
- reviewed how chunks can be bundled between peers
- Michael T. asked about scenario 1 in which node B need to keep its
   previous address, so the text here needs to be tighten
- also, each chunk goes in one packet, and shouldn't be stored until
   there is X many
- agreement was reached between Marushin and Randy/Michael about a max
   of one packet is the barrier for transmission, that multiple chunks
   can be in the same packet.
- Peter discussed v4 or v6 address families
- and that ADD-IP mobility could be addressed with this doc
- look to incorporate into ADD-IP (which dropped two authors, making
   room for these new authors of ADD-IP)
- Randy said this needs a review to see if incorporation is done right
- then ADD-IP ready for WGLC

Randy and Michael - SCTP Threats and Socket                  9:45-9:55
   draft-ietf-tsvwg-sctpthreat-00                             (10 min)

- summarized changes since last version
- ready for WGLC after SCTP AUTH and ADD-IP

   draft-ietf-tsvwg-sctpsocket-13

- addressed changes in new version
- doc ought to float around for a little bit of time before WGLC
- security section needs to be tweaked a bit to account for one more
   thing
- ought to be LC after SCTP AUTH, ADD-IP, Threats

Francois - RSVP Extensions for Emergency Services            9:55-10:05
   draft-lefaucheur-emergency-rsvp-02                         (10 min)

- reviewed changes since last version
- covered how merging rules apply to multicast
- moved informative text and examples to appendix
- doc focuses on network layer capabilities
- and not which flavor (for example, preemption or not) should or
   should not be used
    - these are the examples that are in the appendix, that are each
      spelled out
- Janet Gunn supports this
- Kimberly King supports this

HUM for becoming WG item 100%
HUM against becoming WG item 0%

Phil - Controlled Load                                      10:05-10:25
   draft-briscoe-tsvwg-cl-phb-02                              (20 min)
   draft-briscoe-tsvwg-cl-architecture-03

- review of what these docs are
- showed doc map point to where each of the PCN docs fit into the
   overall architecture
- John Loughley comment - mentioned that he and David Black came up
   with a better way of presenting this arch
    - the marking behavior
    - how bits are marked
    - how things are signaled
    - "other" things
- Phil continued
- Understanding Behavior
- Investigating Algorithm
- Fred commented: what he's observed is that two routers experiencing
   the congestion are on two ends of the same link, and that this
   shouldn't count as double the congestion when measuring.
- Dave McDyson comment: suggested that the docs read like the routing
   of the flows are static, and not dynamic (i.e. changing LSPs).  Make
   it explicit, not implicit.
- Bob clarified that the QOS is decoupled from the routing, to the
   point of this being for the situation in which routing changes the
   path to where there is congestion (or near capacity) already

Bob - Policing and Accountability for Causing Congestion on Borders
   draft-briscoe-tsvwg-re-ecn-tcp-02                         10:25-10:45

- hopes to go to PS
- reviewed updates since last version
- didn't break out the application draft
- mentioned that many off the list continue to break the models
- Dave comment: is there an assumption of forward and reverse paths are
   taking the same path, and Bob said no, the feedback is e2e
- Sally comment: asked if this plans on changing or redefining 3168
   bits
- Bob answered not anymore
- room liked that answer
- Discussed how bottleneck policers don't work
- But that Re-ECN only policies at the ingress, with the egress
   providing feedback to the ingress
- that encryption is a non-issue because RE-ECN is at the network layer
- A version of DOS attack exists, and will be addressed in the next
   version
- David Black comment: copying 3168 header to outer header is fine


   draft-briscoe-tsvwg-re-ecn-border-cheat-01                 (20 min)

- reviewed update since last version
- policing at your domain's border, and not trust neighbor's network to
   policy what you have set to be denied
- this is maybe once every 30 seconds vs. TCP which is every other
   packet
- this is not saying this is how you must to contracts
- goal is not to have SBCs by using this mechanism
- Randy comment: SCTP and multi-homing might produce interesting
   results, Bob agreed
- Lars comment: where does this work go? He thinks this is wider than
   the TSVWG can do (involving the Internet, Routing and Security
   Areas), so a larger discussion needs to start as to where this ought
   to go - and this list needs to be involved in that discussion.
- Matt comment: that some of the authors and contributors are not part
   of the IETF, and remove any IETF list from the cc list on each
   message.
- Aaron comment: asked that Bob come up with a 10 page summary ID
   covering what's in the 90 doc - to get folks interest (who will read
   a 10 page doc, when they likely won't read anything 90 pages).

   draft-briscoe-tsvwg-re-ecn-apps-00
       **this ID didn't get submitted on time**

Francois - RSVP Ext for Admission Control over DS using PCN 10:45-10:55
   draft-lefauchuer-rsvp-ecn-01                               (10 min)

- discussed doc
- differentiates this from other proposals by having the intermediate
   router (domain ingress router) into RESV message to update the PCN
   load caused
- defines a new object
- egress GW determines threshold that is maintained
- new errors codes are required to do this
- this is a new doc, and not asking for WG status yet

Georgios - Resource Unavailability PDB                      10:55-11:05
   draft-karagiannis-ru-pdb-02                                (10 min)

- reviewed idea
- this is without signaling
- idea: using Diffserv to use throughout measurements and DSCP remarks
   to notify resource unavailability
- can be used in combination with e2e ECN

Bruce - A Proposal to Incorporate ECN and PCN in MPLS       11:05-11:15
   draft-davie-ecn-mpls-00                                    (10 min)

- reviewed idea
- this is about marking bits of a particular type (EXP in MPLS)
- ECT assumed throughout core, if assumption is right, packet is
   forwarded, if assumption is wrong, packet is dropped
- use RFC 3270 for all packets marked with another codepoint
- works everywhere except where there is all 8 PHBs (i.e. EXP bits)
   defined already
- wants to be a Standards Track
- on agenda in the MPLS WG, so there remains an open question which WG
   this ID should be in
- Sally believe this should be a WG item, and prefers TSVWG, but Lars
   prefers MPLS because it is regarding MPLS.
- Magnus comment: if PCN is removed, this ID will progress much faster,
   Bruce agreed and can do this


VOTING RESULTS:
   draft-lefaucheur-emergency-rsvp-02
    VOTE for WG item 100%
    VOTE against WG item 0%