Re: email straw poll: literal semantics proposals

[Patrick Stickler, Nokia/Finland, (+358 40) 801 9690, patrick.stickler@nokia.com]


----- Original Message ----- 
From: "ext Jos De_Roo" <jos.deroo.jd@belgium.agfa.com>
To: "Patrick Stickler" <patrick.stickler@nokia.com>
Cc: "w3c-rdfcore-wg" <w3c-rdfcore-wg@w3.org>
Sent: 10 October, 2002 13:21
Subject: Re: email straw poll: literal semantics proposals


> 
> 
> I think I understand what you are saying Patrick
> but I have much less in mind
> a bare literal is *not* denoting anything
> but its own lexical form and I'm convinced that
> that is a global and unambiguous meaning

Then you actually were saying that you and I love the
string "RDF"? I have a hard time believing that. I think
your intended meaning was to denote the Resource Description
Framework, and this illustrates the problem with string-based
semantics of inline literals. People use literals to name
things. Those names just happen to be ambiguous.

> and I agree that looking for more than
> that leads to nonsense
> of course you could then ask where the
> real datatype value is and the answer is
> then that it is not on the semantic web, but
> maybe somewhere inside an engine or so
> the point however is, do we really *always*
> need those *value* designators out there?

Yes, I believe we do, because as I see it the whole
point of RDF is to allow knowledge to flow freely
between arbitrary systems so that even if those
systems do different things based on the knowledge,
it's the same knowledge and the statements mean the
same thing, no matter what the system and without
having to have any extra-RDF knowledge about each
systems presumptions or constraints.

I think what it all boils down to is what we're using RDF
for. If we want to express something like "Both you and I
love the Resource Description Framework" then an interpretation
of "RDF" as a string in 

   You love "RDF" .
   I love "RDF" .

simply doesn't do the job. Because if "RDF" is denoting a string,
then those statements are not saying anything whatsoever about
our love for the Resource Description Framework. They simply
say that you and I both love the string "RDF" (that might
be true, but that's not what was intended to be asserted).

And if we wish to see the kind of synergetic and combinatoric
expansion of knowledge that the SW is envisioned to provide,
we need to prevent ambiguity and conflicts as much as we can
(even if we won't be able to do it entirely) so that if some inference
engine concludes that you and I love the same thing, it should
be globally true, not just true in the context of one particular
system.

> if so, then that could be done with B, C or E

With C, D, or E, yes it can, but that precludes the use of 
inlined literals as contextualized names, which I would
consider to be a substantial loss in usability and 
convenience.

> OK suppose one uses E (as it is already done
> by many people) such as in e.g.
>   :Jenny foo:age _:v . _:v xsd:integer "10".
> then is that "10" really denoting anything
> else but its own lexical form???

Well, this is one benefit of from E evolving into D, that
one does not have to worry about conflicts of interpretation
between inlined literals and literals in datatype property
idioms. With idiom D+F, the problem doesn't exist.

Patrick

Received on Thursday, 10 October 2002 07:35:42 UTC