I think that what Tony suggests would be a very useful thing to do. I don't know why it necessarily implies queries to unknown databases and provenance issues, tho'.
You could imagine this sort of generic query to be very convenient and useful to do even with *known* databases, esp. when you have 20 or more such databases listed on the site (such as we have currently listed on the OpenSkyQuery.net site). A visual examination of all the available datasets could convince the user that these are trustworthy datasets, but they would not have the time or inclination to dig into each archive's metadata to figure out which columns to name in their query.
And even if they did, the query might end up being horrendously complicated. The alternative of a simple query based on UCDs is very attractive, IMO.
Besides, as Tony suggests, trying these types of queries out with the existing (admittedly incomplete) UCDs will really tell us what more we need to add to them and how much can be really achieved with them as they are. I also think it could make for an impressive demonstration of the ease and power of using the VO.
Ani
On Mon, 30 May 2005, Tony Linde wrote:
> Hi Masatoshi,
>
> Thanks for that. This is why I'd like to see the UCD referencing in ADQL
> v1.0: it may not be the best solution but it is the only one available at
> the moment to allow multi-db querying and will allow projects to experiment
> with the concept and report back on what we need to do to deliver a better
> solution via data models.
>
> Once we have some trial results, we can determine the types of quality
> criteria that needs to be recorded in the registry for datasets, and when
> this metadata is included in the dataset resource information, we can
> address the issue of data centres keeping it up to date and relevant.
>
> But we need the software first and for that we need the features in ADQL.
>
> Cheers,
> Tony.
>
> > -----Original Message-----
> > From: Masatoshi OHISHI [mailto:masatoshi.ohishi-at-nao.ac.jp]
> > Sent: 30 May 2005 06:30
> > To: Tony Linde; 'Alberto Micol'
> > Cc: 'voql'
> > Subject: RE: Queries to multiple databases
> >
> > Hi Tony,
> >
> > We, JVO team, discussed on the same issue some years ago.
> >
> > At 15:49 05/05/29 +0100, Tony Linde wrote:
> > >simply because there was no possibility of putting them
> > together. If the >VObs effort is to be really successful, it
> > needs to be more than just making >faster a few things that
> > can already be done: it needs to offer things of a
> > >dramatically new 'conceptual' kind for the astronomer to do.
> >
> > But we haven't experimented the idea, because astronomers
> > (note that all members of JVO are astronomers) don't want,
> > generally speaking, use data with unknown data quality.
> > Scientific results depend on the noise level, positional
> > errors, and so on. Our conclusion was that we need to ask
> > observatory people to guarantee data quality and to provide
> > its index as one of database attribute.
> >
> > At this moment various databases have various data qualities,
> > and it is not recommended to make a blind queries. Instead it
> > is the easiest way for our science to explicitly specify
> > database table names (column names) in our queries.
> >
> > But I think your idea is quite attractive to me, and I think
> > it is good to brush up the idea together with data providers.
> >
> > Cheers,
> >
> > Masatoshi
> >
> >
> >
>
-- Aniruddha R Thakar, Research Scientist, The Sloan Digital Sky Survey Ctr for Astrophys Sci, JHU, 3701 San Martin Dr Baltimore MD 21218-2695 410-516-4850 fax:410-516-5096 thakar-at-jhu.edu www.sdss.jhu.edu/~thakar ----------------------------------------------------------------------- Every generation laughs at the old fashions, but follows religiously the new. [Thoreau]Received on 2005-05-30Z23:32:41