Setting up a registry

[Markus Demleitner]

Hi Simon,

On Thu, Nov 09, 2017 at 12:57:10PM +1100, Simon O'Toole wrote:
> We are interested in setting up a publishing registry for IVOA
> services for our new Data Central system; we have a TAP service and
> will soon have several Simple Cone Search services and an Simple
> Image Access service, plus more services in the pipeline.

Ah -- we should really have a nice and concise page with a couple of
standard answers to this and related questions.

> How do we go about this? We don't want to be re-inventing the
> wheel, so any assistance or pointers would be most welcome!

OAI-PMH as such isn't a terribly complicated protocol, but producing
the records themselves programatically can be a bit involved
depending on how you currently keep your metadata.  So, my
recommendation is to start with that (some publication tools already
do that for you, but that isn't likely to be terribly helpful if
you've written your own software).

There's a short introduction on

http://dc.zah.uni-heidelberg.de/purx/q/enroll/info#write-registry-records-from-scratch

-- feedback on what additional information would be helpful is highly
welcome.

To see if you got it roughly right, use a normal XML schema validator
(if you give schemaLocation).  You can also use purx for that if you
put your piece of VOResource on the web.

purx, http://dc.g-vo.org/purx/q/enroll/custom, in addition is an
alternative to running a full-fledged publishing registry if you
forsee that you'll only have a dozen or so resources.  Have a look at
the docs if you're interested.

As to speaking OAI-PMH yourself, there are several generic tools out
there (e.g., https://www.openarchives.org/pmh/tools/) that may be
helpful in setting up a complete registry.  I'm afraid I can't be
more specific than that since I've never used any of those -- perhaps
someone else can chime in?

There is furthermore some code around for more VO-specific OAI-PMH.
DaCHS has a python implementation that, however, is pretty closely
integrated with both twisted/nevow and DaCHS' metadata management;
take that away and the rest is almost trivial.

What else exists?  CDS, CADC, IRSA, HEASARC, etc: If you have
anything public, please speak up!  I'll collect the responses into
the "RegisterYourServices" page I'd like to create.

        -- Markus