RE: typo in OWL 2 structural specification, not semantics.

Hello,

> -----Original Message-----
> From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org] On Behalf Of Pat Hayes
> Sent: 07 November 2008 22:03
> To: Boris Motik
> Cc: 'Bijan Parsia'; public-owl-wg@w3.org
> Subject: Re: typo in OWL 2 structural specification, not semantics.
> 
> 
> 
> On Nov 7, 2008, at 1:15 PM, Boris Motik wrote:
> 
> > Yes, you're right. I'm sorry - I got confused.
> 
> Yes, I agree, Bijan is right, contrary to my original note.
> 
> Still, this is VERY confusing. I would suggest that the use of
> "positive zero" and the notation "+0", while I am sure well-meant, is
> responsible for much of the difficulty. The whole idea of a "positive
> zero" is incoherent in any case, and the positive/negative contrast
> and the symmetry of the notation both strongly suggest a symmetry in
> the underlying structure, which in fact is quite asymmetric.
> 
> If this convention is already in widespread use, I guess y'all have to
> go with it;  but if not, I'd strongly suggest getting rid of 'positive
> zero' and '+0', and just calling it 'zero' and writing '0'. You know,
> like people have for about a millennium now.
> 

The reason why we have positive and negative zeros is because this is how things work in XML Schema Datatypes 1.1; these are in turn
based on IEEE 754 for the representation of floating-point numbers. Our goal was to be compatible with these specs, so we included
them into OWL. If this were not the case, OWL ontologies could not capture all floating point numbers that can occur in practice,
which some members of the WG considered problematic.

The solution was to "promote" the standard zero to +0, and to introduce a new -0 which is not a real number.

> Another thing that is highly confusing in the text is the use of
> 'equal' having a distinct meaning from 'identical'. If these are not
> synonyms, then 'equal' does not mean =. Boris, I think your first
> response was confused on exactly this issue.
> 
> I think what y'all mean by
> 
>   "+0 is equal to -0 but not identical to it"
> 
> is in fact
> 
> "+0 isn't equal to -0 but they are equivalent as far as arithmetic is
> concerned"
> 

The XML Schema Datatypes 1.1 spec explicitly distinguishes identity from equality. In particular, +0 and -0 are not identical (i.e.,
they are different things in the interpretation domain), but they are equal. Equality is used in OWL mainly for the definition of
facets. For example, the following data range is interpreted as a set of two elements containing +0 and -0:

DatatypeResctriction( owl:realPlus minInclusive "0"^^xsd:integer maxInclusive "0"^^xsd:integer )

The reason for this is that both +0 and -0 are equal, but they are not identical.



In contrast, the following data range is a singleton and contains only +0:

DatatypeResctriction( owl:real minInclusive "0"^^xsd:integer maxInclusive "0"^^xsd:integer )


I'm afraid there is not much we can do if we want to remain compatible with XML Schema Datatypes 1.1 and IEEE 754.

Regards,

	Boris

> Pat
> 
> PS. I presume the ordering intended is
> 
> -INF < any negative real < either zero < any positive real < +INF
> 
> but without any order assigned to {zero, -zero}.  Is this right?
> 
> > I've now reverted the spec into the original condition. While doing
> > this, I've just
> > noticed that we haven't defined the relationship between -0 and +INF
> > and -INF. Therefore, I've added the following condition:
> >
> > * -INF < -0 < +INF
> >
> > Again, sorry for the confusion.
> >
> > Regards,
> >
> > 	Boris
> >
> >> -----Original Message-----
> >> From: Bijan Parsia [mailto:bparsia@cs.man.ac.uk]
> >> Sent: 07 November 2008 19:06
> >> To: Boris Motik
> >> Cc: 'Pat Hayes'; public-owl-wg@w3.org
> >> Subject: Re: typo in OWL 2 structural specification, not semantics.
> >>
> >> On 7 Nov 2008, at 18:55, Boris Motik wrote:
> >>
> >>> Hello,
> >>>
> >>> Strictly speaking, this wasn't a typo because -0 is equal +0. I can
> >>> see, however, that this can confuse people, so I've therefore
> >>> changed the document as you've suggested.
> >>>
> >>> Thanks for this comment!
> >>
> >> Actually, I think this is a bit misleading and may even change the
> >> meaning.
> >>
> >> +0 is a real, thus covered by the first clause (we should switch
> >> these to an ordered list).
> >>
> >> -0 is *not* a real (though it is equal to a real), so, just like -INF
> >> and +INF needs an additional clause.
> >>
> >> (At the moment, I don't see a hook up between = and <. As far as I
> >> can tell, it's possible for -0 = +0 *and* -0 < +0. Indeed, if you
> >> change the definition to +0, then I would say that -0 is not < than
> >> any other element in realPlus and vice versa.)
> >>
> >> For example, with this change, I believe your example no longer
> >> holds. I.e., from that definition -1 and 1 only have +0 between them.
> >>
> >> Cheers,
> >> Bijan.
> >
> >
> >
> >
> 
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 or (650)494 3973
> 40 South Alcaniz St.           (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
> 
> 
> 
> 
>

Received on Friday, 7 November 2008 23:29:37 UTC