Question - History of ivoa onto of astronomical object types

[Markus Demleitner]

Robert,

On Sat, Dec 04, 2021 at 07:21:48AM +0000, Robert Rovetto wrote:
> Does anyone know:- why it did not reach an application stage?
> - why the study or further work did not continue? 
> - Did the ivoa ontology exploration proved to be insufficient,
> partly so, or otherwise (and why)?
> - What caused the current effort (the current IVOA vocabularies),
> rather than continuing with the ontology or a set of ontologies?

At least on this last question I can give my personal view.

Historically, much earlier than the object types came various efforts
to produce a thesaurus in astronomy, and the original activities of
the semantics WG had a significant focus there.  This significantly
informed Vocabularies version 1
(http://ivoa.net/Documents/cover/Vocabularies-20091007.html),
published in 2009 and essentially only dealing with SKOS.

Vocabularies 1 did not (and Vocabularies 2 does not) claim to lock
down the applications for semantic technologies to whatever it lists,
and thus when for Datalink's semantics column (WD from 2013:
http://ivoa.net/documents/DataLink/20131022/) properties were more
appropriate than SKOS' rather loose concepts, and we wanted
"naturally" transitive relationships, we went for RDFS instead.

Further applications of this formalism came up, first in
VOResource 1.1 (most importantly, relationships between VO resources;
think "data collections and services giving access to them"),
replacing flat word lists in XML schema files.  But we found that the
somewhat matter-of-factly "here's some RDF" that was introduced with
the datalink vocabulary left a few things to be desired.

First, client authors needed guidance on how to consume the semantic
resources.  And we needed clear rules for how to add terms to the
vocabularies.  There had been several requests for amending the
datalink vocabulary since something like 2014, and nobody was really
sure who should deal with them, and how.  Well: that is how
Vocabularies 2 came to happen.

As to the question on why there is not a single ontology: Well, I
frankly do not see a use case where it would help having concepts
from datalink ("what sort of information does this link give on
dataset?") together with reference frames ("what sort of celestial or
perhaps planetary grid was used?") together with messenger types
("what sort of particle communicated the signal reported here?").

On the other hand, keeping them separate makes a few things quite a
bit simpler, for instance, because the resources often are trivially
compact (just a few kilobytes), and because a standard can say
relatively simple things like "the refposition attribute takes its
values from the identifiers of the
http://www.ivoa.net/rdf/refposition IVOA vocabulary"; in a unified
ontology, this would, for all I can see, take quite a bit more effort
overall, in particular as regards simple presentation accessible to
plain web browsers.

Does this help a bit?

         -- Markus