[TABLE] Re: TAP1.0 Comments

[Doug Tody]

On Tue, 11 Aug 2009, Alberto Micol wrote:
>
> Regarding the above described symmetry, there might be a problem in par 
> 2.3.5.
>
> In par 2.3.5 there is a table upload example to specify how a multi-position 
> search might work using PQL.
> (btw, it is a pity there is no dedicated paragraph to the multi-position 
> search topic).
>
> The last sentence of 2.3.5, regarding multi-position queries, reads:
> "it is up to the service to pick up suitable position column(s) from the 
> uploaded tables",
>
> What if the uploaded table has got multiple positions (as in the case of a 
> result of a TAP join)?
> Wouldn't that  break the ability to upload a TAP result to another TAP 
> service?
>
> It would be much better to allow the user to specify which coordinate pair to 
> use.

The issue of multi-position queries is really a PQL issue; the main
TAP spec should say no more about the details than necessary, leaving
it up to the discussion of multi-position queries in the PQL spec to
go into the details.  On the other hand, more general functionality
such as table uploads is required in multiple contexts and should be
fully specified in the main TAP spec.

Regarding the specific issue of a table with multiple positions,
the service can have built-in knowledge of what to use for the main
position (as is already the case in a cone search for example).
Alternatively we can use standard metadata to flag the default
position columns to be used, or allow a user to override the defaults
by explicitly specifying the columns to be used.

 	- Doug