Minutes of the IPFIX meeting at IETF 79
About 30 people present
Scribes: Nevil Brownlee & Benoit Claise

Juergen Quittek reported on the IPFIX WG Status.  We have one draft in
the RFC-Editor Queue, one in AD-Followup and one that's just completed
IETF Last Call.  The Configuration Model and Flow Selection drafts are
in final edit, and should be finished shortly.  That leaves two drafts
in our current charter, Structured Data and PSAMP MIB.

Benoit Claise presented an update on the Structured Data draft, which
has generated a lot of mailing list following its WG Last Call.  Most
of the issues have been resolved, but Brian Trammel and Hadriel Kaplan
made further comments.  Discussion continues on the list,

Juergen Quittek presented the PSAMP MID draft.  This only needed
textual conventions for 64-bit unsigned integers and real.  The MIB
Doctors have provided one for Unsigned64, but we will need to create a
new one for IEEE 64-bit floating objects.  Dan Romascanu pointed out
that our Float64 TC should be general, so that other MIBs can use it.

The meeting moved on to presentations of current non-charter work.
Benoit presented the IPFIX Mediation Protocol draft.  This draft
provides a clear overview of how a Mediation Function needs to
behave.  At this stage we have two such functions - Anonymisation
and Flow Selection - with Aggregation (see below) as the third.
The goal for this draft is to make it simpler to creat new Mediator
Functions, each only needs to describe/explain anything that's
not covered in te Mediation Protocol document.
    Brian Trammell presented the Aggregation Draft, pointing out that
although the concept is simple, one cannot avoid interactions
between temporal and spatial aggregation.  An important WG task will
be to check the consistency of all the Mediator related drafts, as
shown in slide 2 of this presentation; in particular, the Flow
Selection draft by no means completes our work on Mediation.
    The meeting expressed consensus for this work, it will be
discussed further on the list.

Brian presented a new draft on "IPFIX IE Doctors."  This seeks to
collect our experience of creating new IEs, and to establish clear
rules and conventions for doing that.  Aamer Akhter commented that his
Performance Measurement draft is an example of how the IE Doctors
document would be used.  Nevil Brownlee commented that this draft is
an important step forward for IPFIX, paving the way for future use of
IPFIX by WGs other than IPFIX itself.  Hadriel Kaplan commented that
we also need a new document, "IPFIX for Dummies."  Dan commented that
although this is clearly important, it needs to done very carefully,
especially for future mission-critical uses.
    There was clear consensus for having this as a WG item; to be
discussed further on the list.

Juergen led a discussion on "should we move the IPFIX base standards
(RFCs 5101, 5102 and 5103) to Draft Standard.  There was consensus
that this would be worthwhile if it corrected all the bugs discovered
to date in the documents, and added explanatory text to make them
easier to understand.  We also need to remove 'IP' from the IPFIX Flow
definition, since that is proving to be a blocking factor or the ITU-T
NGN to reference IPFIX.
    Juergen pointed out that we should organise an interoperability
test to verify that the various IPFIX implementations do interoprate
properly, and proposed that we arrange this somewhere (not at IETF)
around IETF 80 in March 2011.  He asked for anyone interested in
helping to prepare a series of tests to please get involved on the
list.

Five other new drafts were presented:
   Shingo Kashima presented "IEs for the Data Link Layer."
     The idea here is to add more IEs at Layer 2.  Dan pointed out
     that we should re-use whatever we can at L2 (don't re-invent it).
     Several people commented that 5102 already has many IEs at
     layers other than IP (layer 3).

   Shingo Kashima presented "IEs for Short Timer"
     This describes two new IEs that would provide higher resolution
     for Flow Start and End times.  Consensus here was that these
     would certainly be useful, and could be done by simply requesting
     them from IANA.

   Benoit CLaise presented "Exporting MIB Variables"
     This would provide a simple way to read SNMP variables at times
     that relate to flows.  It would require two new IPFIX Set IDs.

   Aamer Akhter presented "Flow Performance Measurement"
     This would provide a way to use IPFIX to transport performance
     data from remote probes (IPFIX Exporters).  Several people
     pointed out that the IPFIX WG should not define any metrics
     or ways to measure them.  That should be done in other WGs.

   Benoit Claise presented "Exporting Application Information."
     The idea here is to develop a set of standard 'Application
     Identifiers' that would be independent of the ports being
     used, and which could be used by all implementers.  This
     will clearly require a lot of discussion if it is to be
     implemented.

We did not call for consensus on these five items, but the first four
of these are potential WG items; they will continue to be discussed on
the list.

The meeting finished at 1135.