adql: prefixes in TAP_SCHEMA.columns

[Markus Demleitner]

Hi DAL,

I'm afraid I must resurrect this somewhat smelly thread about whether
the datatypes in TAP_SCHEMA would be

adql:INTEGER

or just

INTEGER.  

The reason is that RegTAP, which I'd like to push into TCG review
ASAP, prescribes a lot of type names for TAP_SCHEMA, and if we later
decide that the adql: thing wasn't that much of a good idea anyway,
it's going to be a pain sorting out what that'd mean for RegTAP.  So,
excuse me for trying to force some sort of consensus.

What we had so far was Pat arguing in favour of them,

On Thu, Jun 05, 2014 at 02:58:04PM +0200, Marco Molinaro wrote:
> I agree on Mark's suggestion and if this is accepted we'll change
> accordingly our services' TAP_SCHEMA-ta (now we have adql prefixed naming,
> but it doesn't change a bit to the service output if we remove it).

which I'd take to be "slight preference against",

> > On Thu, 5 Jun 2014, Markus Demleitner wrote:
> > > Additionally, TAP_SCHEMA is (presumably) shared between different
> > > languages on the same TAP server (I, for one, am tempted to allow in
> > > subsets of postgres, e.g., for WITH).  It hence would seem unwise to
> > > hard-code ADQL types in TAP_SCHEMA, in particular if they aren't
> > > really ADQL types in the first place.
> > >
> > > Thus, I'd argue we should drop the prefixes and just use "generic"
> > > SQL types (while not disallowing stuffing any old junk in there
> > > ("NETMASK", "BOOLEAN"), as currently consumers of this are humans
> > > anyway).

which I hope conveys I'd like to drop them, and

> 2014-06-05 10:58 GMT+02:00 Mark Taylor <m.b.taylor at bristol.ac.uk>:
> > Based on the state of the argument so far, I'd suggest:
> >
> >    - If possible, put the db type ("DOUBLE", "INTEGER" etc)
> >      as shown in the last column of the table in TAP 2.5
> >    - Don't prefix those types with "adql:"
> >    - But you can put anything else in there if you want, and if
> >      you'd like to slap some characters before a colon at the front,
> >      go ahead

which is again expressing a preference against mandatory prefixes.

I have to say I much like Mark's proposal.

For SQL92 types, I consider the argument of a TAP_SCHEMA shared
between different languages fairly compelling against the prefixes.
For types with an xtype in the TAP type table (POINT, REGION, and
friends), I might be swayed, though I'd still not really expect the
adql: prefix to help a lot there.

So -- does anyone want to champion the adql: prefixes at this point?
In particular, Pat, do you keep up your preference in favour of them?
If nobody steps in, I'd remove them from the RegTAP PR before
submitting it (tomorrow, probably), and I'd propose a clarification
to the TAP specification in the Implementation Notes some time later.

Cheers,

          Markus