Meeting twiki page Re: Radio IG running meeting and ObsCore extension for radio data news

[Markus Demleitner]

Hi,

On Wed, Apr 17, 2024 at 11:53:37AM +0200, Pierre Le Sidaner via dm wrote:
> Having participated extensively in the development of EPN-TAP, it would be
> useful to clearly describe the requirement in each of the axes (temporal,
> spatial, frequency, gravitational, etc.).

The EPN-TAP example nicely illustrates another reason for why
well-defined (i.e., all columns MUST) extensions are better than just
specifying a bunch of columns that you can (or cannot) have in a
single table: Whenever you want to use an extension column to
constrain something, you have to look at the table metadata of *each
individual* EPN-TAP table, which makes all-VO queries using one of
these extensions fairly painful.  I'll mention in passing that from a
publishing tool perspective, managing this pool of extension columns
that publishers can pick and choose from also is suboptimal (but yes:
this is something we do not consider for obscore).

After these experiences, I'd rather nudge EPN-TAP in the direction
I'd like to see for the obscore extensions than model these after
what EPN-TAP does so far.

> The scientific users who will need to access the data and who are not
> necessarily geeks.
>
> By this I mean that joins are not an obvious thing for all researchers.
> However, there's nothing to stop proposing a series of tables that
> correspond to each need and that for each data model we have a view that is
> extracted from the same group of tables.

As I said a few mails back: In my experience people start from sample
queries and don't worry too much about their inner structure.  I'd
actually bet that a solid majority of current ADQL users had a query
with a join in it as their very first query overall: one of the
crossmatch examples that TOPCAT generates so nicely.

So: don't worry about user uptake of the joins.  They've done it for
crossmatching, they'll do it for the obscore extensions, at least as
long as we give them enough examples they can start from.

          -- Markus