Re: error?

On 2002-07-03 21:57, "ext Garret Wilson" <garret@globalmentor.com> wrote:

> 
> Brian,
> 
> ----- Original Message -----
> From: "Brian McBride" <bwm@hplb.hpl.hp.com>
> To: "Franco Salvetti" <franco.salvetti@tiscalinet.it>;
> <www-rdf-comments@w3.org>
> Cc: <boley@informatik.uni-kl.de>
> Sent: Wednesday, July 03, 2002 10:12 AM
> Subject: Re: error?
> 
> 
>> One can think of a range statement as a contraint.  With no range
> statement
>> there is no contraint on the range of rdf:object which is what we want.
> 
> Arrgh. I thought I just spent a huge thread and determined that RDF Schema
> does *not* specify constraints.

There has been a long standing debate about whether RDFS provides
for descriptive or prescriptive semantics -- i.e. whether there
are actual constraints.

I've come to the conclusion that one cannot make any determination
simply based on RDFS alone. It is either, or both.

If there is no typing explicitly defined for an object, then
rdfs:range implies a type. Thus, it acts in a descriptive manner.

If there is typing explicitly defined for an object, then rdfs:range
still implies a type, and if the explicit type of the object
is not equal to or a subclass of the implied type, then an
application is free to raise an error (but need not). In this
way, it acts in a prescriptive manner, as a (potential) constraint.

Likewise for rdfs:domain:

If there is no typing explicitly defined for a subject, then
rdfs:domain implies a type. Thus, it acts in a descriptive manner.

If there is typing explicitly defined for a subject, then rdfs:domain
still implies a type, and if the explicit type of the subject
is not equal to or a subclass of the implicit type, then an
application is free to raise an error (but need not). In this
way, it acts in a prescriptive manner, as a (potential) constraint.

Thus, RDFS is always descriptive, but may in addition, be prescriptive
(introduce potential constraints) but *only* if some application treats
them as constraints. All RDFS is providing is inferences about types.
Whether those inferences conflict with explicitly asserted types is
up to the application to determine and act upon. RDFS does not
require any application to insist that explicit and implicit typing
agree in any way.

So, to a certain extent, to say that RDFS provides constraints or that
RDFS does not provide constraints are both correct, depending on the
nature of the application.

Does that help any?

Cheers,

Patrick

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Thursday, 4 July 2002 02:52:40 UTC