Hi,
A few items to think about with regard to approaching queries to registries.
I don't see much difference between searching a registry and any other searchable database. The main difference, perhaps, is in the metadata you invoke. Thus, I think it makes sense that Registries should ultimately support the general QL that comes out of the VOQL working group.
That said, we may need to adopt something early to work with until we have a VOQL. This experience can provide important input to the VOQL development process.
2. Metadata vs. QL
To ensure that the QL adapts well to our evolving metadata model(s), they need to be independent. That is, the QL syntax must be metadata neutral. It may place requirements on how the metadata is defined, but it should not care what specifically appears left or right of an equal sign.
I believe Keith's strawman reflects this. (This is particularly obvious in the XML version.)
3. Registry Metadata
As you know we have a separate WP for this topic. A general schema for registries is proposed in text form in the RSM document and an XML version based on it is posted at the RWP03 page: http://www.ivoa.net/twiki/bin/view/IVOA/IVOARegWp03. Other proposals are welcome there.
In my mind, what we should concentrate on in RWP03 is the metadata about Resources in particular. For astronomical related information--which certainly should be included as part a resource description (e.g. coverage)--we should leverage off the data models that come out of the DM WG. We should not plan to make up our own just for use with registries.
For example, in my proposed schema (VOResource), it is intended that the "Coverage/Spatial" node would use elements from the Space-Time Coordinates and Regions schema being developed by Arnold Rots et al.
4. XML-based QL vs. SQL vs. XML Query
The motivation for using an XML-based QL is that it is easier to parse than SQL and XQuery. In my mind, this is a critical requirement for the general QL because at some point in the chain (most sensibly at the data provider's site), catalog queries must be transformed into a native format. It is not just a matter of transforming to SQL; the metadata itself must be transformed to the local schema. (A definite requirement of an XML-based QL would be that it is easily transformed into SQL and XQuery.)
This parsing requirement is less critical, I think, for registries, but it could still be. It all depends on how the registry data is actually stored internally (which we don't want to mandate!). Still, in the short term, XML Query is not a bad choice, especially if we plan to use an OO/XML-based metadata model (which I think we should).
Keep in mind: straight SQL is only straight-forward if the schema used in the query matches what is actually in the (relational!) database. If you can't guarantee this, you have to parse the SQL.
cheers,
Ray
Received on 2003-04-30Z19:04:50