Various minor problems in VO-DML repo

[Markus Demleitner]

Hi,

On Thu, Feb 23, 2023 at 10:42:46AM +0100, Laurent Michel wrote:
> > On 23 Feb 2023, at 08:45, Markus Demleitner <msdemlei at ari.uni-heidelberg.de> wrote:
> > And lugging the old minor versions around does have a price tag.  For
> > instance, these minor-versioned URIs may tempt people into putting
> > these into MIVOT's MODEL/@url.   Whether or not that's bad is an
> > interesting question (which perhaps also should be addressed in
> > MIVOT).  If this were XML, it'd be a disaster.
> >
>
> My point of view is that minor-versioned documents are related to
> existing standards, therefore they must keep available on the doc
> repo as long as the (old) standards exists.  Hiding the minored
> versions is somehow like prohibiting using those version. Is that
> what we want?

Yes, I would argue that way.  Since minor versions are intended to be
backwards compatible, I don't think there is a good reason to go back
to older minor versions except for debugging-like tasks, and these
are better served with proper version control where you have comments
on why certain changes were being made.

But I won't quarrel here -- this certainly is one of the lesser
problems we currently have with our VO-DML repo, and we only need to
agree on it when there's going to be a DM update anyway.

> > local copies of those[1] and hence can't easily tell what URIs they
> > may prefer.
> for tne record:
> The MIVOT schema requires MODEL at url not be empty IF PRESENT.
> <MODEL name=“MyModel”/> is valid.
>
> I like the Paul's idea of encouraging the usage of short URLs e.g.
> "MyModel.v1.vom-dml.xml” and letting clients getting them from
> their favorite place (file:// or https://ivoa.net
> <https://ivoa.net/>…) This must be clarified in the MODEL section
> of MIVOT.
>
> Would this help DACHS for dealing with documents with empty uri?

Well, if I don't have to put in the URIs, I'll just look the other
way when there's empty URIs.  But it would certainly be wise to
inquire again why exactly the uri element was made mandatory in
VO-DML.  It sure feels... odd to decide to have a mandatory element
that is in general empty, and we might block some important thing
that the VO-DML authors had in mind.

I can't resist: this is another example where just adding features
without a solid implementational backing makes everyone scratch their
heads later (and that's the best possible outcome).  Which is exactly
why I think we should make MIVOT 1.0 a *whole* lot leaner.

> > So, either the models give that themselves, or I have to provide a
> > local mapping in DaCHS, the latter obviously being ugly and brittle.
>
> I do not follow you, if you need the document to get its URI,  you
> must first have that URI to get the document.

Sure.  But then I have checked it into DaCHS VCS, and by the time
DaCHS reads it'll need to know where it originally came from (when I
need to give MODEL/@uri, that is).  And that I can either take from
the document or from a hand-maintained local mapping.

But if I can safely leave out MODEL/@uri, that's a non-issue *for me*
at this point.

> AS Paul said, you can cross-check the document URI with its content
> by using <name><version> elements.  No need of any <uri> element.

...except for VO-DML files not (yet) in the IVOA repo.  I suspect
that's why people came up with the uri element in the first place.
But at this point I think that's another of our lesser problems.

> A CC-0 label on the top of the XML files would be nice.
> Do you we need errata for this?

Me, I don't think so.

       -- Markus