RE: A certain difficulty

What do you mean by a 'Person'?

Anything and anybody can be described from different points of view.

From the point of view of Social Security a person is something that has a
social security number, an age an employment history.

From the point of view of a bank a person is someone that has a name, an
address and a set of accounts.

If you want to express these different roles in RDF the solution proposed by
Stefan, to define a class for each role and use multiple inheritance to link
it to the resource, is the simplest and most evident solution.

It also provide a handy way of querying for information regarding a given
resource.

A remote db might store different kinds of information regarding a given
resource but you might not want all of them.

You might naturally ask for only some properties to be delivered but in many
cases it would be simpler to say, give me all the information regarding this
object considered as a BankClient (or UKCitizen or Mammal).


What would this be an impedement to?

  titto



> -----Original Message-----
> From: Guha [mailto:guha@epinions-inc.com]
> Sent: 04 March 2000 18:17
> To: Stefan Haustein
> Cc: Tim Berners-Lee; Bill dehOra; www-rdf-interest@w3.org
> Subject: Re: A certain difficulty
> 
> 
> The semantics of Class are different in RDFS and  OO.
> 
> In your proposal, we are unlikely to have a single Class
> "Person". We will probably have a thousand or more different
> versions of "Person", each with its own set of properties.
> 
> I think this will be an impediment ...
> 
> guha
> 
> 
> Stefan Haustein wrote:
> 
> > Tim Berners-Lee wrote:
> > >
> > > No, there *is* something fundamentally different between 
> most distributed OO
> > > systems and frame systems and RDF.  In a frma system, 
> fundamentally, the
> > > designer of an object class defines the set of properties 
> an object may
> > > "have".  This doesn't scale, as it doesn't allow anyone 
> to say anything
> > > about anything: you can only say what the class designer 
> said you could say,
> > > and in some syetms you can only say it if you have write 
> access to the
> > > object. The RDF model in which properties are essentailly 
> first class
> > > objects independent of classes (though constrains can 
> later be expressed) is
> > > fundamentally more weblike, and therefore scalable.
> >
> > I claim that you still could say anything about anything if
> > properties were local to classes in RDF: An instance
> > can have several different types in RDF. You would just need
> > to design a custom class containing the properties you wish,
> > and add the new class to the types of the instance you
> > describe.
> 
> ...
>

Received on Saturday, 4 March 2000 13:51:55 UTC