[TAP] data type for column metadata

[Patrick Dowler]

On 2009-04-01 01:42:36 Miguel Cerviño wrote:
> Maybe I am a bit lost in the discusion but I am a bit afraid  about
> the idea of STC is needed to be implemented any case.

Coordinate system metadata would be optional and in many cases not applicable. 

> I try to work with theoretical services where STC metadata have no
> sense at all.

I recall that STC has a term RELOCATABLE which I think means that there is no 
absolute zero point to the values; this would be suitable for simulations. It 
can be used to describe any kind of value.

> Ok, at this moment I must to include STC metadata in theVOTables, but
> I would
> prefer to have protocols where STC metadata is not mandatory.

Agreed. It is not mandatory. 

The only sense that it is STC is in there at all is that we refer to the STC 
standard (and FITS WCS papers) for the names and definitions of the labels 
one can include to help describe the system. 

> In this sense, I would prefer what Roy points out:
>  > The simple way is NOT to implement astronomical time in TAP!

But that's not really a plausible approach. As Anita pointed out, if we ignore 
these details then users will see all manner of ugly situations and think the 
VO is broken. For example, I get an image from one service and a catalog from 
another and overplot the catalog on the image. The image WCS is in FK5/J2000 
and the catalog is in FK4/B1950. The result is terrible. If the correct 
metadata is there, tools can correctly display this; if they can't do the 
transform, at least they can warn the user that the two inputs have different 
systems. 

This is the easiest example to come up with, but there are many... for example 
the catalog could come from SimDB and have system=RELOCATABLE. You can't 
really do anything with that, but at least tools and users would know how to 
or not to compare it to empirical data.

> Otherwhise, lot of theoretical services that would make use of TAP
> (synthesis models, isocornes,
> models in general needed to we able to perform real data analysis in
> the VO) will be (again)
> outside of the VO-game [except in the Theory interest group that is
> not allowed to produce protocols).

I sincerely hope that this does not happen and welcome all input from the 
theory side, especially if/when we make assumptions rooted in observational 
astronomy.

-- 

Patrick Dowler
Tel/Tél: (250) 363-0044
Canadian Astronomy Data Centre
National Research Council Canada
5071 West Saanich Road
Victoria, BC V9E 2M7

Centre canadien de donnees astronomiques
Conseil national de recherches Canada
5071, chemin West Saanich
Victoria (C.-B.) V9E 2M7