[TAP] sync vs async

Guy Rixon gtr at ast.cam.ac.uk
Fri Mar 6 02:53:17 PST 2009


There seem to be two memes emerging from recently responses.
The first is that parametric query is somehow more fundamental than  
ADQL.
The second is that parametric query is robust while ADQL is fragile. I
don't think either of these ideas are quite right.

We know that ADQL is more general than the parametric query and it  
imposes
less data model than parametric query with POS and SIZE. I think this  
makes
the ADQL service more fundamental.  If a site offered a partial TAP  
service with
just ADQL I could write a parametric-query service that delegated to  
the former,
without having a copy of the database. (This may be a useful tactic  
for NVO if
there arise TAP sites that don't provide the parametric query.)  I  
couldn't do it
the other way round.

As to fragility, there's no reason why an ADQL service should be more  
fragile than
a parametric service. It's the synchronous services that are fragile,  
and a
parametric interface can be asynchronous if it needs to be.

If I wanted to build an application - perhaps part of a portal - that  
did specialized
queries, I would need to build it on asynchronous queries, so that it  
worked
consistently, and I would probably need ADQL. (Consider the case of a  
service
that selected on high B-V OR high V-K; you can't say that in the  
parametric query
because there's no OR operator. Or consider a query that needs joins.)

This is not a request to change TAP. I just ask that we recognize  
that ADQL
and asynchronous queries are important to building everyday systems for
astronomy. They are not baroque toys.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2104 bytes
Desc: not available
URL: <http://www.ivoa.net/pipermail/dal/attachments/20090306/26e9b0b5/attachment-0003.bin>


More information about the dal mailing list