RE: ISSUE 15.24 if or iff - current state - propose to close

I am not particularly opposed to this, but I do find it odd.  We have
taken such excruciating pains to be compatible with RDFS.  And while
RDFS gives us explicit permission to tighten our semantics, why do
something different from both RDFS and DAML?  Uniformity in this case
seems a little spurious, since the other IFF property-properties are
quite different, stated as they are as rdfs:type relations.  And we
still have only-IF charactizations of other property characteristics,
e.g. complementOf (at least as I read
http://www-db.research.bell-labs.com/user/pfps/owl/semantics/rdfs.html#5.1).

I assume we need to change the Reference, 
http://www.w3.org/TR/2002/WD-owl-ref-20021112/, 
which currently states

 "Multiple range restrictions are interpreted as saying that the 
  range of P must be the intersection of all the class expressions"

Under the IFF interpretation, the following holds

 P rdfs:range A
 P rdfs:range B
->
 A = B

Which seems like it might make the construction of consistent
distributed ontologies tougher, yielding unexpected equivalences.

The model theory asserts that if P is an element of
CEXT(S(owls:ObjectProperty)) then its extension is a mapping from IOT
to IOT. Reading that, I might naively assume it to make some assertion
about P's range.  And it does, just not exactly the one I expected.
We know the range of P is a subset of IOT.  We don't know that 'P
rdfs:range IOT'.

So, on the plus side we have 

1. uniformity (which seems spurious.)
2. domain and range treated as prescriptive (more like a programming
   language).

On the negative side we have

1. differs from RDFS and DAML.
2. additional overhead on reasoning systems.
3. greater care needed when trying to add a range or domain
   restriction (e.g. need to make sure there is not already one).
4. domain and range treated as prescriptive (more like a programming
   language).

What I don't have a clear handle on are the use cases that would
inform the selection of either IF or IFF semantics.  And I think this
is probably the most important consideration.

- Mike

-----Original Message-----
From: Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com]
Sent: Monday, November 18, 2002 7:46 AM
To: Jim Hendler; webont
Subject: ISSUE 15.24 if or iff - current state - propose to close




Jim wrote:

> Issue 15.24 if or iff properties
> (chair's note: Is this issue still alive?  If so, can someone propose a
close?)

My belief on the current state:

The RDF semantics gives rdfs:range and rdfs:domain "if" semantics.
The OWL semantics gives both of these and all the other property properties
"iff" semantics.

The RDF semantics explicitly permits OWL to make that change. "Semantic
extensions MAY strengthen the domain and range semantic conditions"

Thus:
- the two semantics together do propose a solution to this issue that meets
(my) preference of uniformity.

The desire for uniformity seemed to be the group consensus, I don't believe
there was a consensus on whether if or iff was better. Given that we have
text with "iff" I propose that we accept that.

Thus:

I propose that Issue 15.24 if or iff is addressed by the OWL Semantics
document.
I propose that we close this issue.

(Note there already are plenty of test cases, I don't think we need any
more).

Jeremy

Received on Monday, 18 November 2002 11:49:34 UTC