This is a brief and provisional report on the discussions at the
"Query and Data Languages" session at the I3/POB meeting Jan 8-12 in
San Diego.  The major results were

   1. Agreement on the semantics of a "minimal" language for 
      data interchange.

   2. Enumeration of some "challenge" problems facing extensions
      to this language.

   3. A decision to pursue these problems further at a workshop in
      the near future together with an assignment of working groups.

   4. Discussion of other issues that may impact on the choice of
      language.


Summary of the meeting
======================

The meeting was organized and summarized by Louiqua Raschid, Jeff
Ullman and me (Peter Buneman).  We started by briefly reviewing our
charge and adding further topics for discussion; we then had brief
presentations from
     Jeff Ullman       (Stanford)
     Tim Finin         (Univ. of Maryland, Baltimore)
     Val Tannen        (Penn)
     Raghu Ramakrishan (Wisconsin)
     V.S. Subrahmanian (Univ. of Maryland)
     Mike Genesereth   (Stanford)

Jeff Ullman proposed that we should address the problem of designing a
minimal language for information interchange -- minimal in the sense
that it would alleviate the problem of the "two-day integration
crisis" by requiring that every agent could be queried in a common
language, even though the language may not express the full functionality of
that agent. 

It was agreed that the design of any "data" language required at least:
      (a) A data model
      (b) A computational capability
      (c) A format for data transmission

There was universal agreement that for a minimal language we need
(a) the relational model and (b) the relational algebra.  The issue
of the syntax of the language and (c) was not regarded as important if
the only goal of the project was to design and implement a minimal
language. That is, people were prepared to use any of the standard
syntaxes for these tasks.

HOWEVER, the syntax of the language was seen as very important when
one wanted to consider language extensions for various purposes.  Two
proposals for syntax dominated the discussion: (1) datalog - because
a simple declarative syntax allows optimization and certain other 
features, and (2) OQL because its extension to more complex types is
already defined.

No immediate agreement was reached on this issue.  It was decided to 
examine the choice of syntax in the light of some "challenge"
issues:
              
	Exchange of rules

	Complex types and semi-structured data

	Abstract types (methods)

	Exchange of Meta-data

Other issues that may effect language design were also discussed:

	Uncertainty. The quality, reliability and completeness
        of data.

        Data movement. When and where should data be materialized?

        Triggers/Active Elements. (View maintenance, constraints
        and inference)

        Temporal characteristics (archival/historical data)

        Security

The Next Steps
==============

There was general agreement that the topic is important and that the
discussion must be continued.  Loiuqa Raschid has offered to do some
of the organization, and individuals have volunteered to do some
preliminary work on the major issues described above:

       
         Rules: David Maier and Mike Genesereth

         Complex Types:  Peter Buneman, Jeff Ullman and Raghu
         Ramakrishnan
     
         Abstract Data Types: Xiaolei Qian and Val Tannen

         Meta Data: V.S. Subrahamanian and Stan Zdonik