do something useful and simple first

[Roy Williams]

I claim that the part of the TAP protocol about ADQL queries is already 
available to our customers. They can go to IPAC or ESO or wherever, read 
about catalog metadata, form their query, get results. There is no added 
advantage to a universal approach unless you can build a single query 
and send it many places. But this is not true, because each query has to 
be built by a human for each catalog -- because the human has to 
understand the meaning of the attributes. Therefore nothing is saved by 
having a unified query interface, because each query still needs to be 
hand-made by a human after reading the metadata. You can't ask catalogs 
blindly for objects with Vmag<18, because most catalogs don't have Vmag! 
Therefore doing complex ADQL queries is a facility that is COMPLEX to 
implement, COMPLEX to standardize, and doesn't give our customers 
anything that they will value.

However, there is a very simple query that can be universally applied 
across may data centers and catalogs -- the cone search. This is because 
every astronomical catalog has RA and Dec. The really useful extension 
to this, that IMHO should be first and center of the TAP effort, is 
"multicone" simply an extension of this, so that multiple cones can be 
sent all at once. This facility is the thing in TAP that is both SIMPLE 
to implement, SIMPLE to standardize, and USEFUL TO CUSTOMERS, and it is 
continually blocked out because first we have to do something both 
complex and of little value.

Roy

-- 

California Institute of Technology
626 395 3670