purl, vocabularies and ivoa.net

[Markus Demleitner]

Hi Semantics,

On Thu, Jun 01, 2017 at 12:18:24AM +0200, Mireille Louys wrote:
> Various vocabularies already exist , in various formats and are used in DAL(
> datalink, RDF, html), Registry (VOResource litterals, CSV), Theory (

Just a correction here: Both Datalink and VOResource are distributed
in Turtle, RDF-X and HTML, and both of them are authored using simple
CSV files.  Versioning and content negotiatiation (i.e., what format
you get when dereferencing the vocabulary URL) are, in both cases,
done according to the W3C 'Best Practice Recipes for Publishing RDF
Vocabularies' at http://www.w3.org/TR/swbp-vocab-pub/.

My personal feeling is that it would be great if this pattern could
be used for all VO (new[1]) vocabularies; I suspect that we'd be
doing our deployers a favour if we could even agree on the
relationships between the terms we'd like them to support; datalink
and VOResource already use

rdfs:subPropertyOf

VOResource in addition

owl:equivalentProperty (and, a bit related, although not a
relationship, owl:DeprecatedProperty)

-- which lets you express hyponymy and synonymy, which I guess should
cover quite a bit of ground.

In case you're curious: 

https://volute.g-vo.org/svn/trunk/projects/dal/DataLink/datalink-terms

and (derived from that and somewhat generalised)

http://volute.g-vo.org/svn/trunk/projects/registry/VOResource/terms

contain the (relatively lightweight) tooling to pull all this off
with a minimum of hassle (if I may say so myself).

         -- Markus

[1] At least theory already uses SKOS; I don't understand enough of
what they do with it to have a clue as to how difficult it would be
to migrate or if that is even technically desirable, because of
course SKOS may be better suited to what they're doing than the
somewhat more formal RDFS or even OWL mechanisms in Datalink and
VOResource.