Re: Proposed tweaks to Annotations

On 6 Oct 2008, at 16:38, Ivan Herman wrote:

> Hi Boris,
>
> as far as I know, the intention of using range/domain as a  
> constraint instead of a
> license to infer typing on the subject or the object is a common  
> mistake people make
> when starting to use OWL.

In general, sure. But people like Alan Rector want to use them  
specially on annotation properties with the intent that tools *other  
than reasoners* will take these as hints. If we syntactically forbid  
them, these folks tend to get quite upset.

> If that is the general usage for annotations, then this is
> simply an erronous usage from OWL point of view;

I don't think it's *erronous*. Alan, for example, isn't expecting a  
*reasoner* to give him an inconsistency or to infer anything (though  
he has no problem if it does). There are *lots* of heurstics that  
knowledge *acquisition* tools use by analyzing the asserted structure  
of the ontology. (This often goes under the heading of "sanctioning".)

Now, I may, personally, not be so happy with this way of doing  
something, but in the absence of a good alternative I don't see that  
this particularly way of handling things is *harmful*.

> ie, I am not sure this is a good enough
> argument to specially deal with it...

It seems a good enough reason for me.

> I may have misunderstood you, though.
>
> (As an aside: don't take me wrong, trying to define constraints  
> like that on the usage
> of a predicate is a legitimate user request and I have met this  
> several time when
> talking to people. It is just that OWL does not look like the right  
> tool for that...)

It provides *some* constraint, e.g., you can't include a disjoint  
element. The alternatives aren't OWL constraint and constraint  
constraint, it's *no* constraint (because you can't say *anything*  
about the annotation properties) and some constraint (at least OWL  
constraints, and with tooling, a bit more).

Cheers,
Bijan.