RE: RDFS bug "A property can have at most one range property"

> From: tarod@softhome.net [mailto:tarod@softhome.net]
[...]
> The problem I wanted to explain is why
> the RDFSchema should use the domain classes as an 
> intersection instead of a
> union, without using DAML, ok? I know DAML is more powerful, 
> but RDFSchema
> should be independent of that, if I want to use RDF to 
> represent an UML
> Model, the intersection in the domain of the property is not 
> as useful as the union... 
> 
>   Where can I find a good point to explain it without using 
> or thinking in
> DAML? 
> 
>   I think the problem lies in RDFSchema being modified 
> according to one RDFSchema-based vocabulary.

The problem with the union approach is that a reasoner (for any vocabulary)
can be presented with unexpected types.  For example:

- Your vocabulary defines the range of a property as SomeClass.

- My vocabulary (developed after yours) defines the range of that property
as SomeOtherClass.

- A third vocabulary (developed after yours, but independently of mine)
defines the range of the same property as YetAnotherClass.

- At some point, a user loads my vocabulary and the third vocabulary into
the same RDFS-aware engine.  The range of the property is now
{SomeOtherClass, YetAnotherClass}.

There are now two options:

1) The range is the conjunction of the two.  In this case, it is possible
that many statements in both vocabularies are illegal, but any inference
from any engine will be well-behaved: nothing that was previously illegal
becomes legal.

2) The range is the disjunction of the two.  In this case, all
previously-legal statements in both vocabularies will remain legal, but
there may be new statements that were previously illegal that now become
legal.  This may cause unexpected behaviour in previously well-defined
vocabularies.

Personally, I prefer the first option --- as a user of these two
vocabularies, I'd much rather know in advance that there might be a problem
than find unexpected snags after I've been using the combination for a
while.

		- Peter

Received on Friday, 16 November 2001 13:07:32 UTC