Three PRs for VOResource 1.3

[Markus Demleitner]

Dear Gilles,

On Tue, Apr 22, 2025 at 01:54:20PM +0200, gilles landais wrote:
> About (2) and (3). I agree for the relation "IsContinuedBy" which
> exists also in DataCite.  But I don't see any equivalent of
> "DiscontinuedResource"  type in DublinCore or in DataCite (not
> surprising).  In a sens the status (OAI-compatible) described in
> (2) makes the job. So what's the benefit of duplicating methods?

What you are suggesting (for other readers: a longer explanation is
below) is to indicate a discontinued resource with a VOResource
status="deleted" and no capabilities, but keep it as a normal
resource as far as OAI-PMH is concerned.

I suppose that combination will rarely result from mistakes, so it
would take that hurdle.

I'm not over-enthusiastic about changing RegTAP to say "keep inactive
and deleted records if X", where X might be "they have relations" or
"have relations of types A, B, or C".  But on the other hand, that's
probably not significantly more trouble than adding a resource type.
In my book, it's still somewhat less obvious, and it may be a bit
easier to publish such records by accident.  Plus, we'll need a
status column in RegTAP's rr.resource.

Let me confess: I *think* keeping the current RegTAP behaviour of
dropping such records is more robust.  But I could certainly be
swayed.

> I think too that removing capabilities is the basis for deletion. But it is
> enough considering EUDAT harvesting?
> (eg: How the registry status is managed in EUDAT?)

Not at all.  EUDAT (b2find) technically works on top of RegTAP
(though they don't know it; they're speaking OAI-PMH), and hence they
(at this point) only see active resources in the first place.

Pushing through deleted records would probably be a good thing
(though I think for what they do a weekly full re-harvest would be
good enough).

> For (unfortunate) deletion in Registry, I would be in favor to:
>
> - removing capabilities.
> ... and update metadata as a tombstone page.  For instance, Data provider
> should update the description with deletion explanation.
> Thus the explanation would be available in pyvo (registry module), and may
> be available in EUDAT(?)
>
> - adding the status='deleted'
>
> - adding the relationship=IsContinuedBy when possible

Well, I don't think we should forbid normal OAI-PMH deletion if
that's what the data providers feel is appropriate; if we wanted to
have some sort of guaranteed persistence, we'd have to define a
registry of last resort that would take over records from dying
registries, and I think the use cases aren't strong enough for that.

But as a solution for the use case "a publisher wants to pull a
resource and point its users to a different one" and with the caveats
above I like your plan about as much as I like mine, and I think both
are about the same amount of work (just in different places).  Hm :-)

Thanks,

           Markus