CURRENT_MEETING_REPORT_

Reported by Peter Furniss/Consultant

Minutes of the Minimal OSI Upper-Layers Working Group (THINOSI)

The THINOSI Working Group met on 15th July at the Amsterdam IETF. Most
of the time was devoted to reviewing the upper-layer cookbook.



Upper-Layer Cookbook

Various issues were discussed in the review of the cookbook.  A point
that came up more than once was how the cookbook should relate to the
parallel work in the OSI regional workshops (the Common Upper-Layer
Requirements Part 3:  Minimal OSI profile (CULR-3)) and in X/Open
(specification of use of the XTI interface for minimal OSI (XTI/mOSI)).

The three (or four) categories of application that the cookbook can
support need to be further clarified, making the distinctions purely on
which OSI facilities are used and not the use of ASN.1.  Examples of
application protocols in each category would be useful.  If possible,
the categories should be linked to those in CULR-3 and XTI/mOSI. This
may not be straightforward due to the different approaches of the three
documents:  OSI-style profile, API definition and implementor's
profile/respecification.  Josee Auber will attempt to compare and
contrast the three approaches in a message to the THINOSI list.

The possibility of the cookbook having a formal statement of compliance
to CULR-3 was discussed.  CULR-3 defines compliance statements by which
another specification (e.g.  the cookbook) can state that its use and
support of the OSI facilities complies with CULR-3.  However, CULR-3 is
being developed in the OSI regional workshops (OIW leading, EWOS
involved) and will probably have to follow the conformance requirements
of the base OSI standards.  Some of the base standards have
over-enthusiastic conformance requirements, which go beyond the cookbook
target, which is ``interworking with conformant implementations.''  This
is especially true for presentation.  Moves are afoot in ISO/IEC to get
this sorted out, but for the time being at least the cookbook should not
be committed to comply with CULR-3.

The cookbook specifies the use of indefinite lengths where possible for
sending the Presentation PCI (this is equivalent to the ``Canonical
Encoding Rules,'' a newly-defined subset of BER). There had been a
suggestion that the opposite choice (Distinguished Encoding
Rules---definite-length throughout) would be a preferable
simplification.  Which is found to be the easier to encode or decode is
considerably a matter of the coding approach taken.  Peter claimed that
for these supporting layers (as distinct from the encoding of
application protocols) the canonical choice was best.

                                   1





Peter will attempt to get a new draft of the cookbook out by the end of
July and submit it for posting as an Internet-Draft.

The eventual status of the cookbook was discussed, and it was believed
that it should be targeted for the standards track, as the specification
of the supporting protocol layers for the relevant applications.


Revision of Charter

Since the charter was written (following the Washington BOF), the
coverage of the cookbook has changed to more than just ``byte-stream''
(although the amount of new text is small), and the charter needs to be
changed to reflect this.  Other possible changes were discussed.  It was
concluded that the intended thinDAP document should stay in the plan,
but the dates may need revision.  Given the nature of the cookbook, it
will be worth considering the production of some very brief ``mapping
specifications'' that would state precisely how particular applications
(Z39.50 for example) would use the cookbook.

Peter Furniss will work up a draft revision and post it to the list.


Over the Fence:  Activities in Other Arenas

CULR-3 has been revised again (June 1993) and it is intended to be
submitted for ballot to become an ISP early next year.

The XTI/mOSI specification will shortly be published by X/Open as a
``Preliminary Specification.''  A Preliminary Specification is valid for
a year, and may change subsequently.


Implementation Plans

Peter Furniss was expecting to start extending the X/osi code to a more
general THINOSI implementation, with XTI/mOSI as the upper interface.
Terry Sullivan is very keen to start something to support Z39.50.


Attendees

Josee Auber              Josee_Auber@hpgnd.grenoble.hp.com
Michael Brescia
David Crocker            dcrocker@mordor.stanford.edu
Walid Dabbous            Walid.Dabbous@sophia.inria.fr
Peter Furniss            p.furniss@ulcc.ac.uk
Brian May                Brian.May@mel.dit.csiro.au
Mark Needleman           mhn@stubbs.ucop.edu
Kamlesh Tewani           ktt@arch2.att.com

                                   2








3