Re: semantics of daml:UnambiguousProperty

On Tue, 23 Jan 2001, Jonathan Borden wrote:

> Dan Brickley wrote:
> >
> >
> > A little question re DAML+OIL semantics. If a property is an unambigous
> > property, does the equality specified by DAML+OIL below hold only at a
> > particular point in time or does it hold across time and change?
> >
> > Eg. if I say some resource X has a foo:contact property
> > whose value is some other resource Y, where Y is "the resource  whose
> > personalMailbox is mailto:lighthouse-keeper@stonyisland.example.com", what
> > does this mean? Might Y differ over time, so long as at any one time there
> > was only a single resource with that property/value pair. What inferences
> > does daml:UnambiguousProperty license in this respect?
> >
> Doesn't this depend entirely on the model? Either this mailBox is or isn't
> unambiguously associated with a single person err resource. If arcs are
> present between two URIs representing two people and this URI then this URI
> is not a good value for this resource. Alternatively when such a value is
> defined, perhaps it should flag an error to enter a second arc.

Problem is that there are two realworld scenarios that we might want to
distinguish. One is where the semantics of some predicate allow, at any
one time, at most one value for that property on any given resource. At
different times, different resources may take that role. The
other scenerio is where we know that role will only ever have a single
occupant. It was a slightly mischievous question in that neither RDF nor
DAML+OIL make any explicit acknowledgement of time, change and other such
nuisances. TimBL suggested I ought to be using daml:UnambiguousProperty
for an app I'm working on, so I need to find out whether this property
has the stronger sense. Hopefully this can be clarified without getting
into the business of defining theories of time and the like.

> This sort of problem comes up in the real world all the time. For example in
> the U.S. a social security number ought be an unambiguous identifier for a
> person, except that they are reused ... it turns out that creating a truly
> time invariant unambiguous property is often a  difficult and expensive
> proposition but does prove good fodder for heated arguments.

Yes, it's a very real problem.

Dan


> Jonathan Borden
> http://www.openhealth.org
>
>

Received on Wednesday, 24 January 2001 05:18:19 UTC