Re: Common RDF parser bug?

[Brian McBride]

> I suggest we distinguish between the resource that is the reified
statement
> and a description of that resource e.g. in an RDF graph.  The resource
> does, in a sense, have all three natural properties of subject, predicate
> and object.  However, that does not mean that an RDF graph has to
represent
> all three properties.  I have recently found it useful to have a partial
> description of a statement.  The particular example I have in mind comes
> from an rdf schema for p3p I have been working on where I have statements
> of the form:
>
>    <site> <collects> _:1 .
>    _:1    <rdf:type> <rdf:Statement> .
>    _:1     <rdf:predicate> <p3p:birthdate> .
>
> which is a way of saying this site collects information about peoples
> birthdays.
>

Seems to me that this approach is asking for trouble.  It looks like it is
trying to sneak in some notion of a universal quantifier through the back
door, but without any real semantics for it.  I suggest that's not a good
idea.

Furthermore, I don't think it's even accurate modeling (in this case,
anyway), even if it may now be allowed.  Most sites (I take it you mean
ordinary web sites) do not contain or collect rdf triples (ie.,
rdf:Statement's) of any variety, so for them the assertion would be false on
its face.  I suggest that a more accurate modeling approach would be to
create a type called perhaps "site-information", which is allowed to be
"about" various resource types.

As a related modeling matter, the set of statements above actually says that
the site collects a ***single*** resource (i.e., the node _:1), rather than
a single ***type*** of resource.  Of course you can write software that will
understand this idiom, but I suggest that it would be better to stay
consistent with the normal use and interpretation of RDF.

Cheers,

Tom P

Received on Saturday, 26 January 2002 01:31:04 UTC