Re: rdfms-resource-semantics

> Why bother defining a special RDF notion of 'resource'?  After all, it
> conflicts with the definition in RFC2396 causing much confusion and a
> special RDF definition of 'resource' is probably unnecessary anyway.

Yes, stop philosophy! Perhaps we should just accept that 
there exists a notion of 'resource' and define a formal 
model of RDF, leaving the 'meaning' of 'resource'
undefined. See below.  
 
> That way, RDF's 4 sets become 'Statements', 'Literals', 'References' and
> 'Properties'.
> 

I will simply adopt the term 'reference' here  (I am not sure if we
need a new word ...) 


What is an RDF model? An RDF model could be defined formally 
as follows: it consists of:

a set of References,
a set of Literals,
a set of Statements,
a set of Properties which is a subset of References

and the functions:  

subject: Statements -> References
object : Statements -> References | Literals
predicate: Statements -> Properties

label_r: References -> URI | {*}
label_l: Literals -> XMLFrag

URI denotes the set of URIs, XMLFrag the set
of well-formed XML fragments and {*} denotes
a one-point set describing the null label
(anonymous Resources). | denotes the disjoint
union of sets.

An anonymous resource is a resource s with
label_r(s)=*. 

Two restrictions:

For all resources r , s with label_r(r) = label_r(s) != * follows: r = s . 
(A named resource has
a unique URI as label.)

For all literals l , k with label_l(l) = label_l(k) follows: l = k. (A literal has a unique XML
fragment as label.) 

Now simply add the further restrictions of M&S's section 5
and we are ready. This is consistant with section 5: there
is an obvious bijection between the two models.

OK, this has to be worked out ..., but: we don't have to
talk about the nature of Resources (References) anymore:
They are simply given by a set together with a function 
'label' that assigns a URI to a resource (reference).
(I think what we learn form the discussion on RDFCore
is that there will be no (consistant) agreement on what 
resources really are).

Different representations of RDF can be defined by
suitable bijections between the given sets.
 
> Just a thought!

not that bad ;-)

regards,
stefan

PS: If someone is really interested in defining a formal
(mathematical) model based on directed labelled graphs,
just let me know ...

Received on Saturday, 2 June 2001 09:24:04 UTC