W3C / MIT Distributed Searching and Indexing Workshop
Negotiation vs. Pre-Agreement (breakout chaired and reported by Mike Heffernan)
Session III, Track B

Items Discussed
*	What classes of things could be agreed on or negotiated?
1) Protocol
2) Query Language
3) Schema
4) Ontology / Taxonomy / Concept Base
5) Transfer Format

*	The concept of z39.50 "lite" was introduced - a profile that is more
extensive than WAIS, but less than the entire z39.50 spec. The mission
of this profile would be to enable distributed searching on the
internet.

*	Relevance ranking and comparison in distributed search has a number
of issues: 
1) The quality of the index engine's relevance ranking scheme can be in
doubt.
2) Need to rationalize ranking across collections where the relevance
of the collections may vary widely.
3) Truly high grade relevance ranking normalization techniques will be
limited by most search engines' reserving their own ranking techniques
as a source of added value (and therefore not being willing to open up
these techniques).
4) As a result, it was recognized that a standard for cross-vendor or
cross-service ranking comparison will be lower grade fidelity than
proprietary techniques used within a service or vendor's product line.

Some short term things that could be addressed
*	Packaging of documents by having a "start here" tag in the HMTL.
*	The Stanford work on standardizing a text search model could serve
as the basis for distributed search pre-agreement.

Some longer term things that could be addressed
*	Content neutral metadata framework that would allow for plugging in
domain specific definitions.
*	RDM was felt to be an emerging "brute force" standard with respect to
metadata encoding (when the metadata takes the form of value-attribute
pairs)