resuming progress on TAP

[Doug Tody]

Hi Andy -

While ADQL is clearly separable (and is largely an externally defined
payload passed in through TAP, just as VOTable is an externally
defined object passed back by TAP), I honestly do not see how to
do something similar with the TAP ParamQuery service operation, any
more than one would try to separate the TAP AdqlQuery operation from
the TAP spec.  Both AdqlQuery and ParamQuery are primary TAP service
operations consisting of parameters and associated semantics defined
by the service interface, with many of the parameters and implied
TAP (table query processing) functionality in common.  If they were
completely decoupled from each other and from the service interface
what you describe might make sense, but they are not.  They have more
in common than otherwise.

I do like the modularity of having AdqlQuery be able to support
alternative query languages (LANG parameter), however ParamQuery
being simpler has no such capability.  If it did I agree it would
make sense to define such a payload separately.  As I have noted in
earlier postings I agree that there are some matters which are common
to all the DAL2 interfaces which can be moved out into a separate
specification, but most of the ParamQuery functionality is directly
related to querying table data and metadata.  It is not a general,
separable abstract query mechanism.

I can't help but notice also that this scheme would have the effect
of once again removing param query from TAP.  The current ADQL query
functionality would not be affected (AdqlQuery remains in the interface
even if ADQL per se is defined separately), the only effect would be
to remove param query from the TAP interface.  For those who want param
query to go away, this is even better than moving it to an appendix.

Once again, as a number of us have pointed out, we are probably 95% of
the way to having a functional TAP spec which meets the requirements
of all the major partners.  Merely reorganizing the document so that
the service operations and related functionality are clearly separated
into separate sections in the document would appear to address most
or all of the concerns which have been raised.

(Keith, you raised many of the same issues and I have not responded
yet as I was taking time to think about it carefully.  But probably
my response here largely presents my view on this already).

 	- Doug



On Thu, 12 Feb 2009, Andy Lawrence at ROE wrote:

> Dear all
>
> I have refrained from jumping in before now because my head is spinning : 
> from the technicalities as well as the pyschodrama ! However, it seems to me 
> that Pat's new proposal - core TAP plus documents specifying "payloads" or 
> "capabilities" is absolutely spot on, with an important proviso.
>
> This proposal is right not just because it will break the logjam but also 
> because it will future-proof TAP. For example if ADQL is revised, the TAP 
> document does not need to change; and if someone invents a new payload or 
> capability, it can be added without changing the TAP document.
>
> However, as many have said it is important that we keep to previous 
> agreements.  As well as the QL-mandatory business, it also means that we must 
> deliver QL and PARAM versions as a single package. To keep this agreement, if 
> we logically decouple them, as I believe we should, then the docs for 
> TAP-core, QL defn and PARAM defn must be delivered  as a single package. QL 
> defn we already have of course.
>
> This means two things. First that the drafts of these docs should be 
> presented for comment at the same time (and very soon please !). Second, that 
> TAP-core and PARAM-defn should be offered to IVOA/TCG/Exec to be approved at 
> the same time.
>
>     andy
>
> ------------------------------------------------------------
> Andy Lawrence : Regius Professor of Astronomy, University of Edinburgh
> Institute for Astronomy,  Royal Observatory, Blackford Hill, Edinburgh
> Email al at roe.ac.uk  Phone +44-(0)131-668-8346
> ---------------------------------------------------------
> Temporary Location : KIPAC, SLAC, Menlo Park, CA, USA
> Email : axl at slac.stanford.edu    Phone : +1-650-926-4828
> -----------------------------------------------------------
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>
>
>
>
>
>
>
>
>