VOResource 1.1: relationship type vocabulary

[Markus Demleitner]

Dear Registry,

On Wed, Sep 28, 2016 at 05:17:56PM -0400, Accomazzi, Alberto wrote:
> Your arguments sound very reasonable, and I agree with your
> proposed approach.  We should avoid breakage until absolutely
> necessary so your soft transition is a good way to go.

Ok -- so, I've deprecated the VOResource 1.0 terms in Volute rev.
3585; have a look at the resulting vocabulary at

http://docs.g-vo.org/vocab-test/relationship_type

Potentially contentious points:

(1) I'm now formally mapping mirror-of to IsIdenticalTo;
semantically, that's evil because IsIdenticalTo is a wider concept
(think, e.g., an organisation that is being re-named but want to keep
the old metadata around).  I'm trying to clarify the situation in the
description of IsIdenticalTo.  I'd suggest we don't sweat the thing;
the current registry doesn't have a single mirror-of relationship,
and common mirrors are covered by interface/mirrorURL in the future.
So, I'd propose to consider mirror-of's presence as standards
baroque ("we can't just pull it").

(2) Contrary to the original plan, the new terms now are IsServiceFor
and IsServedBy, because that seems to be more in line with the style
of the DataCite terms.

(3) I'm deprecating the trivial "related-to".  DataCite doesn't have
a "non-informational" term, and I think that's sound design.
However, contrary to mirror-of, related-to is actually being used
(164 times) in the current registry.  

Here's a list of them:  http://dc.zah.uni-heidelberg.de/__system__/adql/query/form?__nevow_form__=genForm&query=select%20*%20from%20rr.relationship%20where%20relationship_type%3D'related-to'&_TIMEOUT=5&_FORMAT=HTML&submit=Go

My analysis:

(a) Quite a few simply seem to establish resource/publisher
relationships -- I'd say these should go anyway, since we have
dedicated publisher fields; these even allow spefcifying an ivoid.

(b) There are numerous relationships at apparently are supposed to
indicate that services are derived from a certain dataset.  Unless
I'm completely mistaken, I'd say these should really be
IsDerivedFrom.  Or perhaps IsServiceFor.

(c) There's quite a few declaration of relationships between IVOA
standards. Here, I actually doubt the machine-readable declaration of
the relationships between our standards is a manageable use case for
VOResource relationships; yes, it would be cool to derive a "If I change
standard X, I (may) need to change Y"-graph in some way, but really,
I suspect VOResource relationship is a good deal to simplistic for
that.  And of course, standards are essentially learned (I would
hope:-) texts.  Relationships between these (citation) have been the
subject of bibliographic work for ages, and we can't hope to improve
on that within VOResource.

(d) Quite a few related-tos go from tutorials and similar to the
tools they use (for the use case "Give me a tutorial using Aladin on
extragalactic sources").  The current vocabulary suggests using Cites
for this purpose.

I guess that about covers it.  So, I'd say I'm unconcerned by the
plan to nudge the authors of the respective resource records towards
migrating to other terms.  Is someone else  concerned?

Cheers,

        Markus