Re: Reference, beginnings of HP review?

Thanks, Brian, for your thoughtful input. I see your points. Let me 
think it over a day or so, and I'll come back with suggested revisions.

Brian McBride wrote:
> At 10:26 23/01/2003 +0100, Guus Schreiber wrote:
> [...]
>> I guess this is my fault. I agree the wording should be more careful, 
>> maybe something like "One way of looking at a class ..", etc., while 
>> also explaining the intensional view. The problem is that if you want 
>> to explain to people how OWL classes work, the notion of a class as a 
>> set of individuals is an absolute must for making soime sense. How can 
>> we otherwise explain, for example, the notion of a owl:Restriction class?
> Guus,
> First a suggestion, then an explanation of why I'm making a fuss about 
> this.
> Does adopting the following terminology help?
>   o classes have instances, sets have members
>   o an instance of a class is a member of the class extension of the class
> With this we can write things like:
> [[An owl restriction expresses a constraint on the instances of a 
> class.]]  which is simpler than
> [[An owl restriction expresses a constraint on the members of the class 
> extension of a class.]]
> i.e. it allows us to write in a similar style to treating classes as 
> sets, whilst retaining accuracy.  It doesn't work all the time.  
> Sometimes you just have to write explicitly about the class extension of 
> the class.  That's the price of accuracy.
> I'm making a fuss about this because I think accuracy in the specs is 
> important.
> As such I think its a bad thing if the reference document says something 
> different to the semantics document.  I have observed the vast amount of 
> energy that can be wasted interpreting documents that are not clear - 
> viz the recent TAG discussions over what a URI denotes and many past 
> discussions on RDF interest with regards to reification, contexts, etc.  
> If OWL publishes documents that are inconsistent in their description of 
> what a class is, either within themselves or with RDF Schema, then we 
> are creating an opportunity for the pointless waste of similar amounts 
> of energy.
> As I understand it, an owl:Class is not a set, and as such, the 
> reference document should not say that it is, that it can be thought of 
> as if it were, that "One way of looking at a class is as a set" or any 
> other such weasely phrases.  The reason for this is that such 
> distortions may lead people to the wrong conclusions.  For example, if I 
> am encouraged to "look at" classes in the way you suggest, then if I 
> have two classes that have the same members, the I am also encouraged to 
> "look at" any statement about one as also being a statement about the 
> other and that just ain't so.
> Why is it so hard to just to tell it as it is?
> Brian

A. Th. Schreiber, SWI, University of Amsterdam,

Received on Thursday, 23 January 2003 11:54:55 UTC