[topicmapmail] instance-of relationship

[Dan Brickley]

* Lars Marius Garshol <larsga@garshol.priv.no> [2003-09-08 12:34+0200]
> 
> * Daniel Rivers-Moore
> | 
> | As I said in my earlier response to Katharina, "canera" is a
> | subclass of "device" and an instance of "type-of-device". What
> | Katharina needs to do, in my view, is introduce the class
> | "type-of-device", whose instances are classes. 
> 
> She could, but I'm not sure the application actually has a need for
> that class. I guess only Katharina would know that.
> 
> | In a side correspondence, Matthew West has pointed out to me that
> | paradoxes do not arise by allowing classes to be instances of other
> | classes. Even standard set theory allows that. The paradox arises
> | when you allow a class to be an instance of *itself*. 
> 
> He is of course right about that.
> 
> Note that the current data model does not forbid classes that have
> themselves as instances...

In RDF, we work this by not saying that classes are just sets. See 
http://www.w3.org/TR/rdf-mt/#technote for details and further b/g 
on the particular technique adopted.

cheers,

Dan