Re: observation on owl:real

Bijan Parsia wrote:
> On 17 Apr 2009, at 08:41, Dave Reynolds wrote:
> 
>> Axel Polleres wrote:
>>> Since owl:real is disjoint from float and double, it appears that the 
>>> lexical space for owl:real, i.e. those owl:real values lexically 
>>> expressible in RIF, actually just conincides with that of xs:decimal.
>>> So, one could say the lexical space of owl:real is the same as the 
>>> one for xs:decimal?
>>> Given that, I kind of fail to see the need for owl:real at all for 
>>> our purposes... resp., I fail to see where it would NOT coincide with 
>>> xs:decimal.
>>> Opinions?
>>
>> This seems like a bug in the OWL specs.
> 
> It's not.

Fair enough.

>> I had assumed that owl:real would be redefined to be the union of 
>> float, double and decimal (perhaps better named owl:number).
> 
> That isn't actually a helpful type.

It would have potential uses within RIF (gives a name to things that 
count as numerics for xpath) and was given as one argument why RIF 
should adopt owl:real.

> First, we have owl:rational -- owl:real roots rational which is a 
> supertype of decimal.

Agreed. I pointed this out in my follow up message.

> Second, a key point of owl:real is to provide values which are the 
> solutions of equations. We don't need lexical forms for irrationals or 
> even transcendentals in order to be sensitive to them. It's easy to 
> build polynomials with rational coefficients which have solutions only 
> in the irrationals, and are thus unsatisfiable in the rationals or the 
> decimals.

These are not expressible in RIF, the RIF arithmetic operators are 
derived from xf&o and inherit the xf&o restrictions on datatypes of 
arguments and results.

>> Since, as you say, they have actually defined it as disjoint with 
>> float and double I agree it serves no purpose and should be dropped.
> 
> I rather suspect the situation is quite different for RIF as the kinds 
> of equations use very different operators, by and large. (And OWL gets 
> access to them via OWL-Rule combinations).

Quite.

Cheers,
Dave
-- 
Hewlett-Packard Limited
Registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Friday, 17 April 2009 10:09:28 UTC