MINUTES IPPM Meeting @ IETF79
============================

Chair:   Henk Uijterwaal, Matt Zekauskas sent his apologies.
Minutes: Dave Mcdysan and Al Morton took notes, these have been turned into 
         minutes by Henk Uijterwaal
Slides:  http://datatracker.ietf.org/meeting/79/materials.html

0. Administrativia 
   Dave McDysan and Al Morton voluntered as scribes.  Martin Swany and Henk
   Uijterwaal monitored the jabber. 

1. Status of drafts not discussed today (Henk Uijterwaal)
   Draft-ietf-ippm-twamp-session-cntrl and draft-ietf-ippm-twamp-reflect-octets
   were published as RFCs 5938 and 6038. This concludes all currently chartered
   work on TWAMP. 

   Draft-ietf-ippm-spatial-composition is with the RFC editor and will be
   published shortly.  

   Draft-ietf-ippm-reporting-04 is missing a reference to the long term
   reporting draft.  Also, the reordering definitions need to be fixed.  Martin
   Swany plans to do this this week.  As soon as the -05 is there, it can go
   to WGLC.  

   Draft-ietf-ippm-tcp-throughput-tm-07 has been sent to the IESG.  Issues
   were raised and the draft is likely to come back for another review by
   the WG.

   Henk finally said that the items on the WG's agenda are almost done, 
   the group will probably finish its work next summer.  If no new items are
   brought up, the group will be closed by IETF82.
       
2. IPPM standard advancement testing (Al Morton), draft-ietf-ippm-metrictest-02
   Al presented the changes in the 02 version of this draft.  Key questions for
   the WG are:
    - What test methods should we use to when dealing with Infrequent Impairments,
      such as Loss, Duplication, and Reordering?
    - Would a tunnel test with hair-pin turn-around that allows each implementation
      to have both sender and receiver at the same test site (possibly with a network
      emulator included) be acceptable to the Group?
    - Can we set the Equivalence Threshold, or Maximum Error allowed as the sum of
      two terms: An Implementation+Test-Setup dependent Factor, and a Systematic Error
      Factor (set a priori)?  We must remember that the equivalence tests only need to 
      detect inconsistent implementations, and not grind away at implentation accuracy.
    - What specific normalizations will we allow for equivalence evaluated using ADK?
   Al invited others to participate in the tests and for review of the draft. 
   No comments were made, the discussion will continue on the list.

3. Long term reporting draft (Al Morton), draft-ietf-ippm-reporting-metrics-04.txt
   The latest version of this draft was presented.   Most of the 
   the new material here is on Restricted Capacity Metrics.
   A new Type-C convention is proposed for capturing TCP options (port number)
   and congestion control algorithms + other configuration details.
   Need additional thought and proposals on Capacity metrics to capture 
   more useful time scales and variability.
 
   Dave McDysan commented that current IPPM work assumes that all IP packets traverse
   the same set of elements for the same path.  The Internet is changing such that 
   packets with different Protocol Type, port numbers may traverse different elements
   (e.g. an http TCP port 80 web proxy) and hence experience different performance and 
   restricted capacity.  This will have an effect on measurements, it is not necessarily 
   correct anymore that packets sent to port X on a host, will experience the same network 
   conditions as a packet sent to port Y on that host.

4. Coordination statement from ITU-T Study Group 12. (Al Morton)
   ITU-T SG 12 Question 17 is revising Y.1540 to include new Stream Repair Metrics
   and Cpacity Metrics, where the "link" concept will be replaced by the "Basic Section".
   A Basic Section can be either an Exchange Link or a Network Section (composed of
   routers at the boundaries, plus routers, switches, and internal links).
   This may be sufficient to deal with the issues raised with RFC 5136...

5. Comments on RFC5136 (Xiangsong Cui), draft-cui-ippm-rfc5136bis-00
   Xiangsong presented a number of issues he has with this document.  These were discussed. 
   Dave McDysan commented (and he'd put this on the list):
    - Router capacity should include packets per second (pps) in addition to bits per 
      second (bps)
    - Router capacity may also include Protocol Type, port numbers and functions that 
      impact performance (e.g., SA validation)
    - Router capacity can be part of many paths, and the example of a simple linear 
      section should be expanded. For example, if a router has limited pps, then a path with 
      many short packets can impact another path with large packets traversing the same router.
   Al Morton and Joseph Ishac (remotely) commented on the correctness of slide 11.  Henk 
   asked about the list discussion, is the current document sufficient, is a one page 
   clarification useful or should we re-write the document from scratch. Al Morton 
   suggested to start with a 1 page to repair RFC5136.  ITU-T SG-15 may have some outputs 
   that may be useful here as well, parallel work in ITU-T SG12Q17 will also address this 
   problem.  Consensus seems to be that a clarification is necessary.  Phil Chimento, Joseph 
   Ishac, Al Morton and Xiangsong Ciu are interested in working on this.

6. Metrics registry and the way forward.
   Henk Uijterwaal introduced the topic by reviewing what has been said about this in the 
   past.  We have a registry but there is very little usage, there is a proposal to 
   obsolete it (draft-morton-ippm-rfc4148-obsolete).

   Emile Stephan presented his alternative proposal (draft-stephan-ippm-registry-ext).  It 
   can be summarized as follows: users need a clear definition of what the metric parameters 
   are, and thus what property has been measured, but the users don't know which one they 
   want. Proposes to harmonize the various statistics defined in the RFCs with the metric 
   registry.
     - Clear Long-Term Metrics
     - define Sibling Metrics
    - Harmonize statistics definitions
       
   Some discussion started, there was some support for the idea, amongst others from 
   Benoit Claise and Martin Swaney.  There was no consensus on the format: a MIB, XML file,
   or something else.

   Henk asked how to move forward, there was no clear direction here.

7. Round Trip Loss Metrics (Al Morton), draft-morton-ippm-rt-loss-01
   This has been proposed as WG draft to complete set of metrics.  Henk will move
   this question to the list. 

8.  AOB
    * Al Morton asked for comments on draft-morton-bmwg-imix-genome-00.txt, 
    * Al showed a few slides on the loss episode draft from Nick Duffield and asked
      for volunteers.  After the session, Yi Sun from a PRC University volunteered to 
      read the draft and comment...