Re: Vocabulary (ontology) URI vs namespace

On 31/08/12 14:25, Bernard Vatant wrote:
> Stuart Williams in cc who is not subscribed to this list gave me the
> green light to forward his thoughts, copied below.
> In answer to the last remark about the ontology and ontology document
> distinction, Dave Reynolds pointed to the OWL 2 specification
> http://www.w3.org/TR/owl2-syntax/#Ontology_IRI_and_Version_IRI
>
> As Stuart, I'm not sure this specification totally clarifies the
> ontological status of an ontology. As written before, using FRBR to make
> distinct the ontology as an abstract work and its various expressions
> and manifestations seems the way to get out of those misty territories.

Actually I pointed to the next section [1] which I think is entirely 
clear that there is one URI. It may leave the ontological status open 
but it closes the practical status. You use the same URI to:

   o identify the Ontology via   <OI> rdf:type owl:Ontology .
   o reference an ontology via   owl:imports <OI>
   o fetch the ontology document, whether in a browser or in code like
     Bernard's harvester

*If* you believe there is a distinction between the Ontology and the 
Ontology Document and *if* you want to make statements specifically 
about the Ontology Document then you are free to mint a separate URI for 
it (whether introduced via 303s or via rdfs:isDefinedBy) but as far as 
OWL is concerned one URI is necessary and sufficient.

Dave

[1] http://www.w3.org/TR/owl2-syntax/#Ontology_Documents

> -----------------------------
> Stuart Williams wrote :
>
> FWIW I think that there are two consistent views and two sets of
> associated URI patterns.
>
> View 1: An Ontology and the document resource(s) that describe it are
> different things and therefore need different URIs.
> ------
> For so-called "#-ontologies":
>      Namespace Prefix: http://{authority}/{vocabulary-path}#
>      Base URI: http://{authority}/{vocabulary-path}
>
>      Ontology Document generic URI: http://{authority}/{vocabulary-path}
>      Ontology URI: http://{authority}/{vocabulary-path}#
>
>      Base relative terms references    #{term} (but beware of
> same-document references RFC 3996 section 4.4)
>
> For so-called "/-ontologies":
>      Namespace Prefix: http://{authority}/{vocabulary-path}/
>      Base URI: http://{authority}/{vocabulary-path}/
>
>      Ontology Document generic URI: http://{authority}/{vocabulary-path}
>      Ontology URI: http://{authority}/{vocabulary-path}/
>
>      Base relative terms references    {term}
>
> In both cases the namespace URI (which as a namespace has no
> significance - it's just a device to abbreviate a URI) and Ontology URI
> are identical.
> The base URI for #-ontologies matches the ontology document (to make
> relative referencing useful) and for /-ontologies matches the ontology
> itself (again to make relative referencing useful).
>
> View 2: An owl serialisation is merely a representation of an
> Ontology... there is only one (generic) resource here (and its ttl, nt,
> rdf etc. variants)
> ------  ie. the document and the ontology are one and the same thing.
>
> For so-called "#-ontologies":
>      Namespace Prefix: http://{authority}/{vocabulary-path}#
>      Base URI: http://{authority}/{vocabulary-path}
>
>      Ontology Document generic URI: http://{authority}/{vocabulary-path}
>      Ontology URI: http://{authority}/{vocabulary-path}
>
>      Base relative terms references    #{term}
>
> For so-called "/-ontologies":
>      Namespace Prefix: http://{authority}/{vocabulary-path}/
>      Base URI: http://{authority}/{vocabulary-path}/ [1]
> http://{authority}/{vocabulary-path}  [2]
>
>      Ontology Document generic URI: http://{authority}/{vocabulary-path}
>      Ontology URI: http://{authority}/{vocabulary-path}
>
>      Base relative terms references    {term}   [1]
>                                        ./{term} [2]
>
> For the '/-namespace' there is a tension between whether to match the
> base URI with the natural ontology/document URI which makes for slightly
> more awkward relative referencing match it with the namespace name which
> makes for more 'natural' relative referencing.
>
> In either view it is the ontology and document (if you take the position
> of them being different things) identifiers that are crucuially
> important. The base URI and namespace URI are just devices for URI
> shortening and that can be got right (even if not always at first
> attempt - folks can miss that resolving relative references is not
> simple concatenation).
>
> For myself - I have never been quite able to commit (universally) to an
> ontology and its describing document being the same thing. I'm open to
> the possibility... but I haven't found anything in the OWL documents
> that resolves that for me (maybe haven't look hard enough). So I tend to
> be a view #1 /-ontologies sort of a person - misguided though I may be :-).
>
> ------------------------------
>

Received on Friday, 31 August 2012 13:44:15 UTC