VOTable session @ Interop.Moscow

[Bob Mann]

 	Hi folks,

 	I think I accept that Mark and Roy are right that my use case
is better tackled through an identifier column in the VOTable, rather
than an identifier attribute in the <TR> element, but I wanted to follow
up on Rob's comment, even though in so doing I'm straying off this
list's topic somewhat.

On Thu, 21 Sep 2006, Rob Seaman wrote:
> The key point is that the cross-matching 
> should be done with astronomically meaningful IDs, not just indices into 
> databases.

 	This is an important point, and what I had in mind as I wrote
about an identifier in the parent database is what we do here in
Edinburgh - following the SDSS - which is to give each source in our
database tables an identifier which is the HTM pixel code for its position 
at a sufficient fine resolution that it acts as a unique ID for that
source.

 	That satisfies Rob's criterion of being astronomically meaningful,
since it encodes the source position, and should be more future-proof than 
an identifier which is just the row number in the database table, but it's
not a standard, so it's of limited use. We've been doing some work here -
which I hope will result in the deployment of a demo web service quite
soon - with caching cross-match results in a database, and thereby
providing a join index for future queries against the parent databases,
but have had to resort to expressing the match using whatever primary
key has been used in the parent database tables. In some cases, this is
an astronomically meaningful ID - as with the use of high-resolution
HTM IDs - but in some cases it will be just an integer row counter, which
is not very robust.

 	It seems to me that this cross-matching of catalogues lies at
the heart of what the VO is about, and many cross-matching operations
will be sufficiently computationally expensive that you want to cache
the results, rather than have every VO user generate them anew. But
to make persisent cross-matches requires persistent IDs for the rows in 
the catalogues, and I don't see how to achieve this being discussed 
actively in the IVOA, despite the fact that it will be needed if people 
are to do real astronomy with the VO.

 	cheers

 	Bob