Re: Issue-114

On Jun 30, 2008, at 5:31 PM, Bijan Parsia wrote:

> On Jun 30, 2008, at 6:59 PM, Alan Ruttenberg wrote:
>
>> After discussion with Michael, we agreed to narrow this issue as  
>> follows: We propose that the only punning in OWL is against  
>> individuals - that is, anything named in OWL can have an  
>> individual with the same (punned name).
>>
>> This corresponds to what I believe is the commonly requested case,  
>> and simplifies the current situation in which we have narrow  
>> restrictions on certain forms of punning - no object/data property  
>> punning, no class/datatype punning.
>
> This is a very tendentious reading of "simplifies". It's just as  
> sensible, if not more so, to say that this *complicates* the  
> picture because it unmotivatedly adds restrictions that are not  
> necessary by any standard. The simplest would be pun everything or  
> pun nothing. In practice, punning more seems to be less of a burden  
> than punning less.

There are two points of view on this (at least). One relates to the  
complexity of users understanding the implications of punning. I will  
argue that punning individuals to anything is natural and easy to  
understand, which other forms of punning are not. I don't find it  
compelling to cite OWL full as a way of understanding property/class  
punning - In general we have found OWL Full to have confusing  
semantics, and I have understood one of the design ethics of OWL DL  
that things be clearer.

Also, the "simple case" pun anything, is out because of the  
aforementioned exclusions. I don't have confidence that we won't  
accumulate further exclusions over time. So the comparision I see (at  
the moment) "Everything can be punned to an individua"l versus "Most  
things can be punned except the different property types and  
classes / data ranges" (my mistake in earlier message saying datatype).

The restrictions I'm reading from the RDF mapping:

> For each x, at most one of OPE(x), DPE(x), and AP(x) can be defined.
> For each x, at most one of CE(x) and DR(x) can be defined.


> Btw, I forget why class/datatype punning is gone. Presumably, if  
> data/object properties must be distinct, one can always disambiguate.
>
> Cheers,
> Bijan.