Error column in VOResource

[Tony Linde]

We need to work on it but I cannot see the problem with using XMLSchema. But
this does mean having a common data model to work from which is why I've
been pushing to get this done. And then both the dataset header info and the
tabular data are encoded and when you see a <SpectralDomain> or <PixelSize>
element in any data they will always mean the same thing; dataset-specific
metadata can be encoded using custom schema extensions just as we've done in
the registry schemas.

But we're straying into DM and VOTable territory now.

T.

> -----Original Message-----
> From: owner-registry at eso.org [mailto:owner-registry at eso.org] 
> On Behalf Of Clive Page
> Sent: 18 May 2004 16:56
> To: registry at ivoa.net
> Subject: RE: Error column in VOResource
> 
> On Tue, 18 May 2004, Tony Linde wrote:
> 
> > As an example of just how badly VOTable represents 
> metadata, check out 
> > Francois' example of representing GOODS metadata in VOTable:
> >   http://www.ivoa.net/forum/dal/0029.htm
> 
> OK, I see the problem, just don't see any obvious alternative.
> 
> If you attach all the metadata to every field element, it 
> expands from just a few characters "<td> </td>" to a hundred 
> or more.  I grant you that the telecoms companies need our 
> money if they are to avoid more bankruptcies, but we've 
> already expanded a typical table by a factor of 10 in moving 
> from FITS/binary to VOTable/TABLEDATA, and another factor of 
> 10 or more seems to me to be over-generous to them.
> 
> If you rearrange the data by columns, so that you have one 
> copy of the metadata and then a list of vector elements after 
> it, that is much more compact, but requires the data in a 
> typical DBMS or FITS table to be transposed, which is a 
> memory-intensive and cpu-intensive step.
> 
> Or maybe there's some other way that I missed...
> 
> 
> --
> Clive Page
> Dept of Physics & Astronomy,
> University of Leicester,
> Leicester, LE1 7RH,  U.K.
>