Re: PROV-ISSUE-142 (Tlebo): Can roles only be Literals? [Data Model]

On 13/11/2011 23:03, Stian Soiland-Reyes wrote:
> On Thu, Nov 10, 2011 at 15:02, Graham Klyne<GK@ninebynine.org>  wrote:
>
>> (2) "a PROV-DM Literal may be a URI-typed string (with datatype xsd:anyURI),
>> or URI-denoted resource (with datatype rdf:Resource); in either case; such
>> URI has no specific interpretation in the context of PROV-DM."
>
> If it is our own invented datatype, it should be in our namespace(s).
> Although the rdf namespace is also by W3C - I don't think we can
> hijack rdf:Resource and make the URI denote a datatype - in particular
> as this datatype would only be used in *non-RDF*-formats.

+1

> Making our own datatype (in OWL2/Turtle) - something like:
>
> prov:Resource owl:equivalentClass
>    [ rdf:type  rdfs:Datatype;
>      owl:onDatatype  xsd:anyURI;
> ] .
>
> (Of course any occurrences of prov:Resource literals in PROV-DM MUST
> in PROV-O be represented as regular RDF resources and not literals
> typed as ^^prov:Resource.)

In theory, I suppose one could do this, but I don't like it.

Specifically, my understanding of the distinction between literals and names 
(URIs) is that the denotation of literals is effectively defined by the language 
used, where the denotation of names is constrained by expressions in the 
language used.  In RDF terms, you should not try to use RDF to describe the 
interpretation of its own literals.

> But is it easier to not say anything about literal datatypes in
> PROV-DM - as datatypes are different in different serialisations. For
> instance, how would you serialise a xsd:float vs a xsd:double in JSON?

+1

> I understand however that we need to say something about literals in
> PROV-ASN as it is an actual syntax. The text should therefore be
> clarified to talk about PROV-ASN literals rather than PROV-DM
> literals.

+1 (I think - or maybe don't use literals at all - model theoretic treatments of 
FoL that I've seen don't use literals, just language constructs and names for 
everything.)

> If we are going to adopt the Turtle-style for literals, then we should
> also do so for resources. So I would then just use the
> <bracket-syntax>  for resource "literal"s named by URIs and drop the
> rdf:Resource reference.

+1

But I'm wary of doing this.  I Turtle style is adopted, the mapping to XML could 
be more complicated.

> We can then just also say that this is the same literal syntax as in
> Turtle, and also allow numbers without quotes (As they are already
> presented in most PROV-ASN examples)
>
>
> We can rename this section from "literals" to "values" to avoid the
> confusion. So the value can be either a literal or a resource (which
> don't have any further interpretation within PROV-DM)

I think that's probably a very good line to take (kind of analogous to "nodes" 
in RDF).  Then only introduce specific value forms (literals) where they are 
needed to support needed capabilities in the ASN.

#g

Received on Monday, 14 November 2011 11:01:44 UTC