Kona -
On Wed, 25 Apr 2007, Kona Andrews wrote:
>> Of course a simple entry level is important - but we already have cone
>> search and SkyNode, and TAP needs to be a step beyond these.
>
> I would agree that the IVOA "has" conesearch, but SkyNode has not been so
> widely implemented - partly perhaps because it got too tangled up with
> a particular service implementation?
I agree that there were many problems with SkyNode; it was essentially a prototype, and was too complicated, did not separate interface from implementation, and so forth. Nonetheless it was addressing an important problem (distributed queries and cross-match) which we cannot ignore.
> My own view is that on the simplicity/complexity axis, our initial version
> of TAP should be closer to conesearch than to SkyNode - i.e. maximally
> clean and simple. That way, we have more chance of getting compliant
> services and clients widely implemented.
Sure; at the most basic level, TAP is essentially a generalized cone search. But it will need to be production grade as we desperately need robust applications which are more than toys (hence large queries cannot be simply ignored). As we have already agreed in our architecture, most of the complexities of distributed queries and cross-matching can be moved to the cross-match portal.
These are still mostly meta-discussions; very little of substance has been said here on the TAP design.