Session III - Format Negotiation / Standardization - Track C

Session Chair: Dave Eichmann, U. Houston - Clear Lake
Scribe: Ken Weiss, U. California - Davis

This session was strongly populated by content providers (LoC, UC
Davis, etc.) and search engine providers.  The discussion was hence
fairly consensus driven on the appropriateness (even necessity) of
structured information regarding both individual artifacts (documents)
and collections (aggregations of documents).

The first 3/4 of the session involved a rather freewheeling discussion
concerning the nature of Z39.50 (both in data/meta-data structure and
the protocols supporting the negotiation of that structure) and
alternatives such as Harvest and its SOIF format.  There was strong
advocacy of Z39.50 as a mature, fielded standard that handles
virtually everything that is under discussion in the workshop (at
least from a content provider's perspective), e.g.,
	* Services available,
	* Attribute list, and
	* Query language.
The issue of establishing the identity of a collection was also
raised, along with the observation that a URC can reference a stored
metadata description which in turn may or may not reference a
registered set.

Concerning formal structure vs. freeform descriptions, there were two
basic perspectives, the ease of automatically generating/interpreting
formal structures and the human readability of freeform descriptions.
Given the distinction in intended consumer communities, there is room
for both in potential standarization efforts.  One key aspect of this
discussion is the perception that exhaustive search requires
structure.  However, there was a definite sense that beginning with a
freeform description still allows convergence towards structure as
deployment / adoption progresses.

Other observations:
	* We try to catalog everything on the net - but shouldn't
	* Early deployment usually wins - but must be free to users to
	  succeed
	* Standards currently address the client system - not the user:
		syntax, spoken language, record structure
	* Negotiation should occur in stages
		common among servers
		specific to a server

======================

Short term goals
----------------

* Work towards a general consensus that negotiation is good

* Develop Z39.50 gateways to AltaVista, et. al.

* Support negotiation of character set and language

* Education on the nature of Z39.50 and its benefits

* Metadata must exist

Middle term goals
-----------------

* Bulk uploading of metadata to indexing server

* Better facilities for metadata set registration

* Persistent user profiles

Long term goals
---------------

* Agent technology to handle negotiation
	creation of "Wizards"