Resend - Outlook crashed while this was on the way - apologies if it double
posts...
A metadata query should go to the Registry. If the registry has coarse granularity then it can harvest the metadata from the resource but if it is fine-grained then it can answer all metadata queries itself.
The user, via the query interface or workflow composer, can then construct a data query (in a high level ADQL) to be run on the chosen resources. If the resources have common UCDs then the query could be entirely constructed of UCDs. The workflow is then turned into tasks within a job where each task will target a specific resource with a specific ADQL query.
Whether the task ADQL contains column names or UCDs is not too important but I would prefer the translation to happen at the resource end. We probably need to cater for both.
Cheers,
Tony.
> -----Original Message-----
> From: owner-voql-at-eso.org [mailto:owner-voql-at-eso.org] On
> Behalf Of Roy Williams
> Sent: 07 October 2003 21:32
> To: Wil O'Mullane; Martin Hill
> Cc: 'voql'
> Subject: Re: A Working document has been submitted
>
>
> Perhaps there is a mis-understanding between metadata queries
> and data queries.
>
> First is the metadata query, asking which columns are
> appropriate. This would be done in Xquery or ADQL (as
> previously discussed), and it returns a small result. It
> could be in terms of UCD matching, or even Google type
> matching in the column descriptions.
>
> Once we know which columns we want, and what they mean, we
> can issu a data query. A data query must specify exactly what
> is being done, and may return a lot of data. It would be in
> SQL or something very like it (eg ADQL).
>
> Roy
>
> --------
> Caltech Center for Advanced Computing Research
> roy-at-cacr.caltech.edu 626 395 3670
>
Received on 2003-10-07Z23:25:17