About VO registry

[Markus Demleitner]

Dear Jorge, hi Costas,

On Thu, May 23, 2019 at 05:35:31PM +0300, Kostandin Vrioni wrote:
> I have been using the Euro-VO Registry for many years to register
> some VO services of my institution (IAC, Tenerife, Spain) and I am
> very happy the application, so many thanks for your work.     
> We are having an issue with the discovery of our services
> registered in the Euro-VO Registry and maybe you can help me.
> Although the services seems to work correctly, for example (SIAP):       
>           
> ivo://iacvo/IAC80
>        ivo://iacvo/PSZ1y2     
> I'm not been able to find them in most part of the VO discovery
> tools; I've tried these:

Interesting.  Let's see: The OAI-PMH interface at the EuroVO Registry
actually delivers the record *now* when asked about it
(http://registry.euro-vo.org/oai.jsp?verb=GetRecord&metadataPrefix=ivo_vor&identifier=ivo://iacvo/IAC80)
and even includes it when asked what changed when that record last
changed:
http://registry.euro-vo.org/oai.jsp?verb=ListRecords&metadataPrefix=ivo_vor&from=2018-12-21&until=2018-12-22&set=ivo_managed.

Meanwhile, none of my registry mirrors had seen that record in what
they harvested over the last six month or so (which, of course, the
do for the Euro-VO publishing registry, too).

I didn't bring all that together  and told the Heidelberg mirror to
just do a full re-harvest of the EuroVO publishing registry.  And lo
and behold, the records are now there.  They're now available in
TOPCAT (Aladin will take a bit longer, probably until tomorrow).

I can't quite say what happened (Theresa, who holds a lot more
history than I do, might be able to provide more insight), but I
suspect the records made it into the publishing registry quite a bit
after the datestamp and thus were lost in incremental harvesting.
The weak part of that theory is that the records are old enough to
have been part of the initial harvest.  Why they were missing there I
really can't say.

Costas, I think the best way to fix the problem on all searchable
registries (without requiring their operators to do a full
re-harvest) would be to bump the datestamp on these (and perhaps
other records that may share their fate) to today.  That way,
incremental harvesting will bring them back without further
intervention.

Thanks,

           Markus