Discovering Data Collections, implementation

[Markus Demleitner]

Dear Registry, Dear DAL,

This is also going to DAL in hopes that the operators of major TAP
(and perhaps multi-resource SIAP) services read it; if you're just
worried about protocols, you can stop reading here.

The proposed endorsed note on discovering data collections,
http://ivoa.net/documents/Notes/discovercollections/20161111/,
is a scheme to replace GloTS for the discovery of TAP tables and have
a mechanism in its place that is also applicable whenever you have
multiple data collections in, say, a SIAP or SSAP service.

There's an RFC page for it at
http://wiki.ivoa.net/twiki/bin/view/IVOA/Discovercoll11RFC

On the implementation side, Mark Taylor now put in something very
close to what I hope to eventually see as the next generation of TAP
table discovery; you can enable it in the TAP window, TAP -> Service
Discovery -> Reg prototype; get the latest jar at 
ftp://andromeda.star.bris.ac.uk/pub/star/topcat/pre/topcat-full.jar

It would be great if you could check if your service is in there as
you expect it (i.e., with the tables, whether declared directly in
the TAP record or an auxiliary capability, represented).

If not, please take the opportunity to review the note or talk
directly to me.  The scheme isn't hard, but you need to have

* tablesets declared, in either the main record or per-dataset (for
  most current TAP services typcially SCS) records; if you're already
  running TAP, that's simple, you just take what's exposed at the
  tables endpoint.
* if you have per-dataset records, these need to have the auxiliary
  capability (which is stereotypical per dataset), and the
  relationship to the main service

One or two of these conditions currently aren't fulfilled for roughly
30 services, among them some fairly large ones; these are shown with
a "?" instead of the number of tables.

Before going ahead and dunning TCG members to cast their votes, it
would be great if at least these large providers could see if they
find major showstoppers.  Let's try to make this "essentially ready"
by the time it goes to become an endorsed note (or fix it if there
are major obstacles).

Thanks,

             Markus

PS: Oh, and if you don't like what's proposed, it would be great if
you had alternative ideas; we'd be honored to discuss them over on
the Registry mailing list...

PPS: Particular thanks go to Mark Taylor for patiently implementing
this despite the fact that GloTS apparently works well enough for his
users...