W3C home > Mailing lists > Public > www-rdf-interest@w3.org > October 2004

Re: Ways to restrict the properties a class may have?

From: Jan Algermissen <jalgermissen@topicmapping.com>
Date: Fri, 29 Oct 2004 15:53:20 +0200
Message-ID: <41824B50.6070700@topicmapping.com>
To: Benja Fallenstein <b.fallenstein@gmx.de>
CC: algermissen@acm.org, "www-rdf-interest@w3.org" <www-rdf-interest@w3.org>



Benja Fallenstein wrote:

> Jan Algermissen wrote:
>
>> I know about the various ways to constrain individual properties, but
>> I seem to be unable to find out if there is a standard way of 
>> restricting
>> the set of properties instances of a given class may 'have'.
>> (Like relational tables form classes by grouping attributes)
>
>
> No -- whyever would you want to do that? 

To express a constraint for enforcing data integrity.

> That would be trying to deprive RDF of one of its best features, the 
> ability to extend others' schemas. 

I agree. OTH, checking that an employee cannot have a numberOfWindows 
property is not a bad thing after all :-)

>
>
> Let's say you're modelling a class Employee. Do you believe that you 
> should be able to declare which properties an Employee can have? Well, 
> an employee is almost certainly a subclass of Person; and if I create 
> a x:starsign property on Persons, who are you to decide that Employees 
> cannot have this property? 

I am using RDF for data modeling and data integration purposes and 
expressing constraits is clearly a value
that I want to deliver to my client. Regarding the semantic Web, I agree 
with you.

Anyway, it might be better to have software detect a disagreement of 
various authors than to silently
ignore it.

>
>
> Now, there's a slighly better case if you're modelling an abstract 
> kind of entity, such as the relationship between an employee and their 
> employer (because an employee can be employed by more than one 
> employer, with different salaries etc). But even this kind of resource 
> may have a property like rdfs:seeAlso or rdfs:comment, as well as 
> similar properties created by third parties, and you shouldn't be able 
> to prevent this by saying "here I have ordained the list of all 
> properties that instances of this class may possibly have." 


Well, true....but go to data modelers and they (usually) just tell you 
that the properties 'make up' the class...

Jan

> Cheers,
> - Benja
>
Received on Friday, 29 October 2004 13:51:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:10 GMT