Re: Meaning of URIRefs (new test case, comments on Concepts draft)

> >But you're free to use other pages.   You could use 
> >   http://www.whitehouse.gov/president/#Bush and be committing to some
> >lesser statements, like that in his 2002 State of the Union Address,
> >Bush outlines three priorities.
> >
> Ah, but this then forces me to use a different URI reference, with no 
> way of stating that the resource denoted by this URI reference is the 
> same resource that is denoted by the first URI reference.  How can I 
> then communicate with anyone who uses the first URI reference to refer 
> to George-Bush-the-lesser?

You're right that I've provided no way to directly equate your URIs
with those of someone using definitions you don't agree with.  I
haven't prevented such a mechanism, and we could define one, but it
would have to keep its "hostile" URIs in strings.

The right way to safely merge nodes, though, when there is no spirit
of cooperation, is through inference.  If two documents both identify
Bush in some unambigous manner (such as through
owl:InverseFunctionalProperties), then we can infer they are the same
person/thing without any inclusion or even awareness of the URIs each
other is using.  No communication arrangement is necessary; the facts of
the matter are sufficient.

(The inference will have to involve natural language understanding
(and thus perhaps people in the loop) if there's really no common
underlying ontologies about things like people's names, dates,
locations, etc.)

Obviously we're better off if we can have extensive URI/ontology
reuse, and that should be generally acceptable if people write their
definitions carefully.

> >I expect you, being more-than-competent in the field would chose
> >instead some more-formal description involving an ontology of the US
> >Govt or of Heads-of-State, and identifying Bush perhaps as the 43rd
> >head of the state of the US.   That kind of definition would have very
> >few unwanted commitments.
> >
> Au contraire.
> 
> At the very minimum, it would identify George-Bush-the-lesser as the 
> 43rd president of the United States.  Suppose that I did not want to 
> commit to this fact?  I would be forced to use a yet different URI 
> reference, that I could not relate to any of the above URI references.

I think there's a tradeoff between how thoroughly you describe
something and how able someone else is to recognize what you have
described.  Two descriptions of him as the 43rd president, named
such-and-such, born on such-and-such a date, etc, are going to be
easily merged.  A description of him as a US President with the last
name "Bush" will be more problematic.  (I suspect any description that
does not use something world-spanning like an inverse functional
property will not really allow one to conclude the objects are equal.)

You could also use a null definition and be back in the "floating
semantics" game, where your readers will have to guess what you meant.

> To take a closer-to-home example, suppose I believe that you are, in 
> fact, an alien?  How can I express this in a way that lets others know 
> that I am talking about  the Sandro Hawke that is linked to from 
> http://logicerror.com/sandroHawke?

I don't see the problem here, unless you disagree with something
AaronSW says about me on that page.  A better URIRef for me would be
   http://www.w3.org/2000/01/sw/people#sandro
If you use that, you're committing to the fact that I have certain web
pages and email addresses and even (oddly) a friend.  (Sort of; I'm
not sure how many of those predicates actually have a non-null
definition.)  You're also committing to similar facts about some other
people.  It's still not a very good definition of me (it wasn't
written to be one), but it's not too bad.

If all that's okay, you can go ahead and say I'm an alien using that
URIRef.  (If you do, though, I might be entitled to use alien
super-logic and counter your arguments before you even make them.  Why
don't we give it a try?  :-)  

    -- sandro

Received on Thursday, 24 October 2002 21:51:49 UTC