Rethink the Constraint-based search Query from Registry interface

[Paul Harrison]

OK - I will come straight out with it - I think that "2.1 
Constraint-based Search Query", the ADQL/XPath based registry query 
interface is an ugly compromise that suits no-body.

I think that it should be reconsidered before it goes into a standard.

The basic problem is that it uses so little of ADQL and XPath that these 
standards should not really be associated with this interface; It would 
be better to define this small subset independently from the other 
standards, so that the registry interface definition would not have to 
be updated to modify the restrictions whenever these other standards change.

Only the Where clause of the ADQL is used, and then there are some 
restrictions on what to do with the meaningless attributes in the 
context of a registry. There are other restrictions that are not 
mentioned in the document - cross match constrains, one of the 
motivating forces behind inventing ADQL, are also meaningless of course. 
So the relationship with ADQL is tenuous to say the least, and to force 
this relationship in the WSDL carries quite a bit of unwanted baggage.

The XPath predicates are not allowed, and only the abbreviated syntax on 
the child axis is allowed, so again the restrictions form a larger 
subset of XPath than what is allowed!

When stripped down, the Constraint based query interface is really only 
a slightly extended version of the keyword search interface, that allows 
better specification of whether exactly what keyword or attribute you 
are searching for along with a more sophisticated boolean expressions 
for combining constraints. I think that the two should be combined into 
a single simple interface with a human readable, simple query expresion 
language that has essentially the same semantics as the current proposal.


-- 
Paul Harrison
ESO Garching
www.eso.org