On Fri, Sep 26, 2003 at 09:24:51PM +0100, Matthew Dovey wrote:
> It is a design criterion of SRW to have a single query language (unlike
> Z39.50 supporting different query types). The intent is that CQL after a
> few revisions will be expressively rich enough for any IR query.
> However, I see GIR more as a framework for GRID based IR than a
> protocol, so feel that GIR should support multiple query languages -
> locking down to specific query languages being an implementation issue,
> i.e. someone inheriting the GIR portTypes to build a GIR system would
> restrict that system to one (or more) query languages.
This discussion may not be relevant to the SRW list, but I don't see
the distinction between the interoperability goals of GIR and
SRW/Z39.50, i.e. that one is a framework and the other a protocol.
They operate within two different environments, but I'd think that
even in GIR we usually want to avoid a situation where a client cannot
query any arbitrary GIR service merely because some support CQL and
others support other query languages. If anything, since one of the
goals of GIR is to distribute IR across many layers of heterogeneous
GIR services, I think interoperability would be even more important
here than in the client-server model of Z39.50. In that sense, the
distillation of Z39.50 into SRW and CQL really helps us. (Apparently
though, at this most recent meeting, some of the escape doors
"popular" in Z39.50 have now been added to SRW!)