Re: log:forSome/#rdfms-identity-anon-resources

Dan Connolly wrote:
> 
> Frank Manola wrote:
> > Dan Connolly wrote:
> [...]
> > > [[[
> > > 4. A person is between a rock and a hard place.
> > > [...]
> > >
> > > Following is the KIF representation:
> > >
> > > (exists ((?x person) (?y rock) (?z place) (?w hard))
> > >         (and (betw ?y ?z ?x) (attr ?z ?w)))
> > > ]]]
> > >
> > > --        Conceptual Graph Examples
> > > http://www.bestweb.net/~sowa/cg/cgexampw.htm#Ex_4
> > > Thu, 22 Mar 2001 01:45:12 GMT
> > > linked from http://www.bestweb.net/~sowa/cg/
> > > linked from http://www.w3.org/DesignIssues/CG#[2]
> >
> > Yes, that was the interpretation I'd been assuming as well (see my
> > message
> > http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jun/0326.html)
> 
> Ah... good... then we're agreed, on the essential point.

I'm not quite sure about that (as usual, the devil is in the details),
so I'm going to follow up a bit on your bookseller example.  This may be
where my apology about barging in should really have come, since I think
I'm now bringing the discussion back to the point where Pat replied to
your bookseller example (I'll try to use different words though!)

> 
> [...]
> 
> >  So I don't understand how queries come into
> > this.
> 
> Then never mind; somebody was convinced by the bookseller
> scenario about the need for existentials in RDF, so I thought
> I'd try it here. But I don't need the WG to agree with
> the bookseller/query scenario; only with the interpretation
> of triples as (exists (?a1 ?a2 ...) (and (p1 s1 ?a1) ...) ).
> 

The need for existentials is one thing;  how they are interpreted is
another.  Take your example:

> In detail: let's suppose you're buying a book.
> You transmit, to the bookseller, a description
> of the book you want to buy; roughly,
> "it's by Barnsley and it's called
> Fractals Everywhere. I want it in
> hardback, tomorrow."
> 
> Formally, in n-triples:
> 
> ---8<---
> _:g0 <http://purl.org/dc/elements/1.1/title> "Fractals Everywhere" .
> _:g0 <http://purl.org/dc/elements/1.1/creator> "Barnsley" .
> _:g0 <http://booksellers.example/vocab#binding>
> <http://booksellers.example/vocab#hardback> .
> _:g0     <http://booksellers.example/vocab#shipping>
> <http://booksellers.example/vocab#nextDay> .
> ---8<---

The problem is that in the interpretation of RDF triples we've been
discussing, if I ran across these n-triples on the Web, I should
interpret them, not as a request, but as an assertion that "there is a
book, which I'm calling -:g0, by Barnsley, with title "Fractals
Everywhere", in hardback, with shipping "nextDay" " (ignore for now
whether it makes any sense to talk about shipping in this context).  If
I used Sergey's URI generation algorithm on this RDF, I'd substitute
<http://skolem.example#432oj34oij2o3ijo23j> for _:g0, so the assertion
would now be:  "there is a book, which I'm calling
<http://skolem.example#432oj34oij2o3ijo23j>, by Barnsley, with title
"Fractals Everywhere", in hardback, with shipping "nextDay";  but the
essence of the assertion would be the same, only the name I'm now using
to refer to the book has been changed.

The issue of context (or query vs. assertion) comes up in your example
when you send that RDF to the bookseller.  If the interpretation of the
RDF as an assertion continues to hold, this would simply be a statement
to the bookseller that "there is a book...", to which the bookseller
might reply "that's nice" or "so what?".  In this interpretation, it
doesn't much matter whether, in what you send to the bookseller, you
refer to the book locally as _:g0 or
<http://skolem.example#432oj34oij2o3ijo23j>.

If the RDF is to mean a statement of requirements (a query), then the
interpretation is at least slightly different, and may be more
substantially different. The slightly different interpretation could be
something like "there is a book, which I'm calling _:g0, by Barnsley,
with title "Fractals Everywhere", in hardback, with shipping "nextDay"
", do you have one that matches?  Changing _:g0 to the Skolem doesn't
change the essence of this situation;  in this case the interpretation
of the RDF would be "there is a book, which I'm calling
<http://skolem.example#432oj34oij2o3ijo23j>, by Barnsley, ..." do you
have one that matches?  The bookseller may have a book that *s/he* calls
<http://booksRus.example/inv2001-06-25#item342323> that matches the rest
of the description, and the problem is to decide whether 

<http://booksRus.example/inv2001-06-25#item342323> is the same book as
_:g0 on the one hand, or <http://skolem.example#432oj34oij2o3ijo23j> on
the other.

[The situation is the same in a purer logic notation:  the request might
say "there exists a book x, ..." and the bookseller's "fact base" might
say "there exists a book y, ..." and the problem would be matching x and
y;   even if the bookseller's assertion was also "there exists a book x,
..." you would be unwise to assume that your x and the bookseller's x
were necessarily the same, without doing the rest of the feature
comparison, since the variable names come from different
namespaces/contexts.] 

In the more substantially different interpretation, you're not really
making an assertion about the existence of the book on your side at all,
you're asking the question "*is there* is a book, by Barnsley, with
title "Fractals Everywhere", in hardback, with shipping "nextDay"? [and,
if so, bind its identifier to _:g0].  Here, you really are interpreting
_:g0 explicitly as a true variable, rather than a local name, and asking
some process to determine bindings for it.   And I would say this isn't
quite the same as saying that _:g0 is an existentially quantified
variable in a logical assertion.  

In either case, you want what you're calling the book to be interpreted
differently in a query, because there you need to indicate which parts
of the description of the book are purely local/arbitrary and need not
match exactly, and which parts are essential descriptions of the book
you want.  I.e., in the case of the query, you want to distinguish "what
I call the book" (which may differ between the requestor and the
bookseller, and thus need not match) from the title, author, etc., which
must match to satisfy the request (and a true variable, rather than a
local name, is a much more explicit way to indicate that).  The reason I
think queries complicate the situation is that the additional meaning
you have in mind for the existentials in queries is beyond the strictly
logical interpretation of existentials in assertions.

--Frank

-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-8752

Received on Friday, 29 June 2001 14:19:43 UTC