Sebastian Hammer wrote:

> The question is whether we can specify a sufficiently powerful
> multi-database mechanism without throwing the simplicity of the protocol
> overboard.

Is it really so complicated?

We would need a databaseName parameter (optional, repeatable) in the query,
databaseName element in <RecordData> within <record> in <searchRetrieveResponse>
and some hit count information.

And semantics.  But the semantics are well-developed for Z39.50, and would be much
simpler in an SRW context. They could be extracted/modified from:
http://lcweb.loc.gov/z3950/agency/wisdom/pdu-multi-db-support.html, and
http://lcweb.loc.gov/z3950/agency/clarify/db-status.html
or look at:
http://www.niso.org/standards/resources/Z39-50-200x.pdf
3.2.2.1.2 "2. Multi Database searching" and 3.2.2.1.7  "When query is not
supported for a database", etc.

--Ray