CURRENT_MEETING_REPORT_


This was an organizational meeting for the ODV group.  The first meeting
was a large one.  (The attendance list is given at the end of this
message.)  It discussed primarily general issues.  There was a brief
meeting of a smaller group of people in the evening, to explore doing
some actual implementation work.

The first meeting discussed primarily the question of whether there
should be an ODV protocol at all.  In addition, issues raised by the
cisco patent application were discussed.  A major part of the meeting
was taken for a presentation by Jose Garcia-Luna of some research of
his.

Many people would like there to be only one routing protocol.  This has
obvious advantages in terms of interoperability.  Since OSPF is now at
the RFC stage, it has a head start in terms of IETF politics.  The
question is whether it makes sense to work on another protocol.  Raising
this issue is about as far as one can go.  The IETF charter does not
make it possible to prevent a group of people from working on a
protocol.  So we didn't vote on the question of whether work should
proceed.  But I will note here that many people were very sceptical.

Part of the problem is that it is difficult to prove in any unambiguous
way what protocol is the best way in the long run.  Jose Garcia-Luna's
simulations attempted to compare SPF and distance vector approaches, but
the routing algorithms simulated were not based on the best
implementations of either approach.  As part of the work of this group,
we are going to try to get the resources to carry this work further.
(This may actually be a more important activity than designing another
protocol.)  My feeling is that routing is still an unsolved problem.  It
is unrealistic to expect progress in this area to stop, leaving some
current protocol as "the answer" for all time.

In response to the concern about extra protocols, I believe we are going
to proceed as follows:

   o Some subset of us will attempt to bring a description of IGRP to
     the stage of an RFC. The whole issue of whether it should be
     considered an alternative to OSPF is one for those who care about
     such issues to negotiate with the IAB. I do not plan to involve
     myself in that.  My feeling is that enough people in the community
     are using IGRP that it at least makes sense to have a generally
     available document that describes it.  If network politics make it
     impossible to issue it as an RFC, it will be available as a Rutgers
     University technical report.
   o We will pursue Jose's work.  This is more of an attempt to advance
     the state of the art than to produce an immediate competitor to
     OSPF. I believe it will be one to two years before anything
     concrete comes out of this.  This work will include analysis as
     well as protocol design.  We will try to avoid producing a protocol
     unless it worth doing.

There was a discussion about the implications of the IGRP patent
application.  There was a very strong feeling against an IETF-sponsored
protocol that is tied up in patent rights.  Some caveats:

   o There is precedent for a protocol that involves a patent.  The
     privacy taskforce is advocating an approach to Email that requires
     a license from RSA, Inc.
   o The concern was primarily that it should be possible to distribute
     public-domain implementation through mechanisms such as the BSD
     tape, for use by recipients.  This does not necessarily rule out
     all licensing.  This request would be consistent with allowing
     internal use by recipients of the BSD tape, but licensing any
     products based on it.

We took a straw poll about licensing.  27 people objected to a protocol
that involved a license.  3 saw no problem with it.  12 abstained.
However it is not entirely clear what this vote meant.  My best guess,
based on a small number of conversations with individuals, is that the
27 people might be satisfied with a public-domain implementation that
allowed free use, but required a license for incorporation into a
product.  At any rate, I believe that the committee will do everything
possible to make any new protocol it designs unencumbered.  This means
that it will not be based directly upon IGRP. To the extent that it
shares the same roots as IGRP, there may still be similarities.  However
we will try to make sure that we have sources in the literature
predating IGRP for any mechanisms that we share with IGRP. Obviously the
attempt to produce an RFC for IGRP will not adhere to these guidelines.

Jose Garcia-Luna's presentation was based on a published paper, so I
don't intend to describe it here.  (I have managed to lose my copy of
the paper.  Hopefully Jose will send a citation to the list.)


ATTENDEES

 Almquist, Philip  Hinden, Bob
 Arnold, Susan     Honig, Jeffrey C.
 Bagnall, Doug     Huston, Geoff
 Baker, Fred       Karels, Mike
 Berggreen, Art    Knowles, Steve
 Borman, David     Lear, Eliot
 Burgan, Jeffrey   Little, Mike
 Catlett, Charlie  Long, Dan
 Chiappa, Noel     Merritt, Don
 Chinoy, Bilal     Miller, David
 Choy, Joseph H.   Opalka, Zbigniew
 Collins, Mike     Pleasant, Mel
 Coltun, Rob       Rosenstein, Mark
 Elz, Robert       Rutenberg, Vald
 Farinacci, Dino   Schiller, Jeff
 Fidler, Mike      Sheridan, Jim
 Forster, Jim      Vaudreuil, Greg
 Fuller, Vince     Veach, Ross
 Garcia-Luna, Jose Willis, Steven
 Gross, Phill      Yasaki, Brian
 Hays, Ken         Youssef, Mary
 Hedrick, Charles