Z39.50 and the Web
Breakout Session at July 2000 ZIG Meeting
Preliminary Agenda
Revised July 8
- Z39.50 URL definition: does it need revision; what functionality should it support?
- Distributed directory of Z39.50 servers.
See
ServerList DTD
- Record Source Schema and related issues.
(
Synopsis)
- GRS vs. XML
- how to import a GRS schema definition
- grs capability to distinguish the cases of "element not there", "element empty",
"element not requested", and the "diagnostic", for which there are no xml
analogies.
- GRS-1 can carry metadata which does not pertain to the record as a static
database object, but which relates to the context of the query which retrieved it.
The best example is probably the HitVector which identifies the parts of a record
which match a query. Together with the variant mechanism, this allows the client
to request only the parts of the record which match the query. If Z39.50 is to be
used to search large full-text documents or similar, this is an important feature.
- GRS-1 can be contrasted to XML in that GRS is a retrieval record syntax while
XML is a generic document format. As such GRS imposes more constraints on
retrieval data than does XML. Some things that GRS-1 defines at a structural
level will require Profiling for XML. Without that profiling, clients applications
will have fewer options for analysing and re-using data from different
applications. But by requiring "proifiling" we lose much of what xml promises to
begin with: interoperability, portability, off-the-shelf support, ...
- On the other hand, XML can associate Style Sheets with specific DTDs (and,
presumably, with schemas). These style sheets may be used to control the display
of records in dynamic ways. In a sense this is fundamentally different from the
Z39.50 model, which talks about structured information and tries to allow client
applications to make the most possible sense to retrieved records. The objective
is not necessarily to display the data - in fact, Z39.50 says little or nothing about
displaying data. The immediate presentation of retrieval records to the user is one
possible application, but not the only one. Others include updating local
databases (such as bibliographic catalogues or bibliographies), merging data from
different sources to present a single result set to the user (a primary objective of
Z39.50). Style sheets that are oriented towards display really offer little assistance
here. It is essential that the underlying structure of the data is constrained
sufficiently that commonality of metadata elements may be recognised and dealt
with. This doesn't necessarily mean that XML is a bad thing, but if implementors
jump into it carelessly, XML-structured records will have little more application
than SUTRS or HTML. If the ZIG is to embrace XML, then it must lay down the
ground rules for best practices, or else forfeit a good 50% of what makes Z39.50
universally useful in the first place (structured retrieval).
- Z39.50 Transport. TCP? HTTP? SOAP?
- XML Query
- RDF