CURRENT_MEETING_REPORT_


Reported by Bob Stewart/Xyplex

SHR Minutes

Agenda


   o Is this trip really necessary?
   o Principles of operation?
      -  Architectural purity.
      -  Interoperability.
      -  Cost/benefit.
   o Definition of Special-purpose Host?

      -  By overall function?
          * Terminal server.
          * File server.
          * Toaster.
      -  By subfunctions?
          * Network self load.
          * Programming interface.
          * Terminals (character devices).
          * Files (FTP, NFS, etc.).
          * Network management client.
          * Network management agent (e.g., bridge, router).

   o RFC Format and organization?
   o Specific issues?
      -  IP fragment reassembly from < 576 byte fragments.
      -  TCP efficiency (e.g., Jacobson retransmission in a ROM).
      -  Source routing.
   o Contributors?
      -  Analyses from vendors of example systems.
      -  RFC section authors.


The Agenda's first question was ``Is this trip really necessary?''  The
consensus was affirmative.  We need some clarifications, the contention
was over how far they should go and what form they should take.

On the question ``Principles of operation?'', we generally agreed that
interoperability is the primary goal.  George Conant of Xyplex suggested
that our first concern should be maintaining the strength of
requirements whose intention is to protect the network from misbehaving
hosts.  Bound by this principle, we can then apply some cost/benefit
analysis to ``musts'' required for architectural purity or use by
wizards under unusual conditions.

Considerable discussion and disagreement did not result in an answer to
``Definition of Special-purpose Host?''  Although subject to debate, the

                                   1






majority seemed to think along the lines of recognizing optional
functional areas, such as an open programming interface, limited
application protocols (such as just Telnet), and so on.  This points to
the consideration that ``special-purpose host'' simply means a host that
isn't general purpose.

The question of ``RFC Format and organization?''  should have included
``Degree of Specificity?''.  Stev Knowles of FTP Software led the charge
for stating principles and omitting specifics.  He was not alone (as if
that matters), but the majority believed we must be more specific.
David Jordan of Emulex spoke for organization by system type (for
example, terminal server, file server).  The strongest consensus was to
organize around the RFC 1022/1023 ``musts'', examining each in the light
of hosts with useful application subsets.

On ``Specific issues?'', the consensus on IP reassembly was ``Shut up
and do it.''  Source routing was less clear.  The idea of keeping the
requirements to forward and to respond on the reverse path but weaken
the requirement to originate a source route had noticeable support.
Issues such as TCP efficiency appear subject to the rule of not hurting
the network while allowing space for knowing exactly how your limited
TCP user (such as a ROM) will use TCP.

The answer to ``Contributors?''  was:


   o Stev Knowles will supply a statement of principle.  We will then
     judge whether we are done.
   o Bill Westfield of cisco and Robert Elz of the University of
     Melbourne will each supply an analysis of the ``musts'' which might
     be subject to weakening in special cases.
   o David Jordan will propose an all-inclusive list of special host
     types.


Discussion was lively and varied, with many valued participants other
than the few mentioned above.  Discussion will continue on the mailing
list.  According to the (unchanged by the way) charter, the next
milestone is a draft document by the end of October, for review at the
December IETF meeting.  The above contributions will provide the text of
that document.

Attendees


Philip Almquist          almquist@jessica.stanford.edu
Larry Brandt             lbrandt@sparta.com
Asheem Chandna           ac0@mtuxo.att.com
Anthony Chung            anthony@hls.com
Paul Ciarfella           ciarfella@levers.enet.dec.com
George Conant            geconant@eng.zyplex.com
Robert Elz               kre@munnari.oz.au

                                   2






Richard Fox              sytek!rfox@sun.com
Karen Frisa              karen@kinetics.com
Martin Gross             gross@polaris.dca.mil
Peter Hayden             hayden@levers.enet.dec.com
Ruei-Hsin Hsiao          nac::hsiao
Ole Jacobsen             ole@csli.stanford.edu
David Jordan             ...jordan@emulex.com
Ajay Kachrani            kachrani%regent.dec@decwrl.dec.com
Michael Karels           karels@berkeley.edu
Frank Kastenholz         kasten@europa.interlan.com
Stev Knowles             stev@ftp.com
Sam Lam
Paul Langille            quiver::langille@decwrl.dec.com
John LoVerso             john@loverso.leom.ma.us
Yoni Malachi             malachi@polya.stanford.edu
Keith McCloghrie         kzm@his.com
David Miller             dtm@ulana.mitre.org
Craig Partridge          craig@nnsc.nsf.net
Stephanie Price          cmcvax!price@hub.ucsb.edu
Michael Reilly           reilly@nsl.dec.com
Craig Smelser
Bob Stewart
 rlstewart@eng.xyplex.com
Bill Townsend            townsend@xylogics.com
Justin Walker            justin@apple.com
Jonathan Wenocur         jhw@shiva.com
Bill Westfield           billw@cisco.com



                                   3