RE: Proposed agenda

> From: Jonathan Rees [mailto:jar@creativecommons.org]
> . . . .
> I don't get this. What makes IRs special? Is it that maybe we
> aren't talking about a class? But I hear statements from all
> group members of the form "x is not an IR" and while they may
> be mysterious or controversial, they don't seem ill-formed or
> semantically problematic.

No, it's fine to think of IRs as being a class. More explanation
below . . .

> > . . .  No amount of
> > discussion of whether an IR has mass, has a dc:title, has a
> > number of pages, has a license, has spelling errors, etc.,
> > will help.
>
>
> I don't understand this either. If you're saying it's not an
> ontology problem, like that of coming up with the most fruitful
> way of talking about chemical reactions or credit card
> accounts, then what sort of problem is this?

There are two problems. One indeed is an ontology
problem: how to define "information resource" in a way that is most
useful to web architecture. I do think it would be helpful for
the TAG to adopt a better definion than the one that is currently
in the AWWW document.

But aside from the ontology problem there is the problem of what
that definition means to semantic web architecture -- how it is
used -- and this is the issue of identity and reference.

It is natural to want to separate these two problems
and focus first on the (simpler?) ontology problem,
but I don't think it works to do that, because without first
having a clearer idea of what that definition will mean in the
architecture, attempts to better define "information resource"
lead straght into a tar pit.

For example, instead of asking:

        "Can an IR have mass?"

the question would be better stated as:

        "In semantic web architecture, can a resource be both an IR
        and also have mass?"

(Answer: Sure.) The question has been rephrased in two important
ways:

 - The context is semantic web architecture, because that's the context
in which we need to ask such questions.

 - The question acknowledges that in semantic web architecture, a
resource can be viewed in more than one way: its characteristics
can simultaneously match more than one thing. So depending on the
ontological definition of IR, a particular resource might very
well be considered both an IR and something that has mass.

The point is that a clearer understanding of identity and
reference in semantic web architecture dramatically changes the
ontological debate: suddenly we don't need to worry as much about
whether some hypothetical thing *is* an IR. Instead we can focus
the ontological discussion on the characteristics of an IR that
matter to semantic web architecture, while recognizing that a
resource can have IR characteristics in addition to having other
characteristics.

> >
> > > So I propose to take up your question: What are the
> > > characteristics of an IR? Or more broadly, what *might* be,
> > > what *has to be*, what
> > > *cannot be* the characteristics of an IR?
> >
> >
> > If you believe what I've been saying about how resource
> > identity works in semantic web architecture, then the answer
> > is simple. Assuming C is the set of core assertions for the
> > definition of IR:
> >
> >  Q: What *might* be an IR?
> >  A: Anything whose core assertions are logically consistent with C.
> >
> >  Q: What *has to be* an IR?
> >  A: Anything whose core assertions subsume C.
> >
> >  Q: What *cannot be* an IR?
> >  A: Anything whose core assertions are logically inconsistent with C.
>
>
> OK. (Although this seems to contradict your assertion that in
> order to define IR we need to get into questions of identity
> and reference.) Tell me what you think C is and we'll have
> something to talk about.

The point of the Q&A above is to illustrate the fact that,
regardless of which definition of IR is chosen, a clearer notion
of identity and reference makes these questions much simpler to
answer.

> Well... you have said one characteristic of an IR (an assertion
> in C) is that an IR is a function, and I have balked at the
> idea that functions have authors.

That's a perfect example of what I mean. It isn't a question of
whether a *function* can have an author. It's a question of
whether a *resource* can be both a function and something that
has an author, i.e., whether the core assertions of that resource
subsume the core assertions for "function" and the core
assertions for "thing that has an author".

> If you're trying to play some trick, or apply a modification to
> the normative RDF semantics,

No trick, and AFAIK I'm not trying to change the normative RDF
semantics.

> . . .
> We're just trying to come up with some rules to guide decisions
> about what's OK to say (and infer) in RDF. We seem to agree
> that we don't want to be allowed to say that an IR has mass, .
> . . .

Again, I think that's the wrong question. It isn't a question of
whether an *IR* has mass -- an IR does not intrisically have mass
-- but whether a *resource* can be both an IR *and* something
that has mass, and the answer to that is very much the issue of
resource identity and reference.



David Booth, Ph.D.
HP Software
+1 617 629 8881 office  |  dbooth@hp.com
http://www.hp.com/go/software

Statements made herein represent the views of the author and do not necessarily represent the official views of HP unless explicitly so stated.

Received on Monday, 23 June 2008 21:35:52 UTC