Re: Chainsaw?

Hello Graham,

Graham Klyne wrote:

>
> [...] I'm trying to pursue the idea of contexts as a tool for structuring complex descriptions.
> (If
> that is a bad idea, I'm hoping someone will say, and explain why.)

Bad idea! :-) I believe my proposal can model in a more natural way what you want to do.

>
> I want to treat [FordEscort] as a resource that has various properties that
> describe that model of car, and, by reference, to have those properties
> apply to my car.
>
> Maybe [FordEscort] should be an rdf:Class?  I'm not quite comfortable with
> the idea that the model may be a class (hence object of rdf:type) and also
> an "ordinary" resource in its own right with lots of properties relating to
> some physical entity.

I think this discussion is closely related to an OIL discussion some weeks ago. There, we tried to
model 'Carnivore' as a subclass of 'animal' with some special properties.

Modelling FordEscort as a class doesn't seem strange to me at all. Personally, I like to see a class
as a set of elements that all share common characteristics. I like to explicitly state those
characteristics.However, the current RDF Schema spec doesn't support fixed property values of a
class. That's why I proposed to extend the 'domain' property.

I one states that Ford Escort is the set of cars with properties (I concatenate your subjects and
predicates)

* bodystyle = Hatchback
* engine_fueltype = Petrol

then, just by instantiating the class Ford Escort, you immedialely get all these standard
properties, applied to your instance! With the 'extended version' of domain, you would get:

<rdfs:Class ID="Car"/>

<rdfs:Property ID="bodyStyle">
<rdfs:domain rdf:resource="Car"/>
<rdfs:Property>

<rdfs:Property ID="engine_fueltype">
<rdfs:domain rdf:resource="Car"/>
<rdfs:Property>

<rdfs:Class ID="Ford_Escort">
<rdfs:subClassOf rdf:resource="Car"/>
<s:bodyStyle> Hatchback </s:bodyStyle>
<s:engine_fueltype> Petrol </s:engine_fueltype>
</rdfs:Class>

With this approach, you can still see ford Escort as a resource that you can describe with
additioanl properties if you want. So it is a class and also an "ordinary" resource. I think this
solutions suits your needs.

If you really want to make the distinction: in context a , statement x is true and in context b it
isn't, then I think you have to stick to bags of statements.

Greetings,

Tom.