W3C home > Mailing lists > Public > public-sws-ig@w3.org > November 2005

Re: Where are the semantics in the semantic Web?

From: Harry Halpin <hhalpin@ibiblio.org>
Date: Fri, 25 Nov 2005 09:29:53 -1000
Message-ID: <43876631.8040407@ibiblio.org>
To: Bijan Parsia <bparsia@isr.umd.edu>
Cc: public-sws-ig@w3.org

Quick response and then no more - and I end on what I think people are
trying to say in a constructive manner...

As for Bijan's confusing denotational/operational semantics, see Page 1
of preface to Formal Programming Languages by Glynn Winskel or
Wikipedia. Operational semantica (from Plotkin) has evaluation and
execution relations specified by rules directed by syntax, the
denotational semantics (from Scott) uses rich mathematical models ala
partial orders, least fixed points, I mean - they are not in opposition,
but different styles with different *uses*  (operational is good at
type-checking!) and vastly different *models* (STM machines vs. classic
lattices for examples). Now, this doesn't mean their unbridgeable - they
are often complementary - one could imagine a operational semantics
defined over types that proved equivalence via DL denotational models. 
(and I would have thought most people here were process calculi kinda
people too...so this all must seem quite primitive) See
http://en.wikipedia.org/wiki/Formal_semantics_of_programming_languages

However - that is a good idea which I was trying to phrase earlier[1]
but failed. Basically,
it seems the best way to make everyone happy would be to figure out a
way to compose WS using a programming-language that bound
OWL/RDF/FOL/"Your Favorite Interoperable Model or Even Bunch of Tags
Here" in XML - and *bound* that explicitly to XML, and let programming
languages aware of this binding do the convesion to model(like RDF
possibly)/composition/discovery in the background as much as possible. .
I think making this work would require locking Pat Hayes and Phil Wadler
in a room together for about a month :)

I mean, I sort of thought that's why people were working on binding
here. Without that, then telling more complicated stories like above
becomes impossible. Instead, I'm seeing lots of the usual RDF vs. XML
arguments.

Is the problem that people feel the binding to RDF/OWL is too
restrictive? Is there something else you want to bind to, or you just
want to bind to anything? Or is that you want to not do any binding and
do it all with XSD? If people want to do binding in XML and have no RDF
whatsoever involved, why not XLink?
                   
                -harry

[1] http://www.w3.org/2005/04/FSWS/Submissions/30/tenpoints.html
Bijan Parsia wrote:

>
> Can I just say that I'm a snarky mood and this thread really pushes
> my  snark buttons. I've been trying to edit it out, but some slips
> through.
>
> On Nov 25, 2005, at 1:28 PM, Harry Halpin wrote:
>
>> Bijan,
>> There's a difference between formal semantics as used in denotational
>> semantics and the operational semantics ala XQuery/XPath.
>
>
> I believe that *is* a denotational semantics, in the sense of:
>     http://citeseer.ist.psu.edu/context/8571/0
>
> But perhaps I'm wrong.
>
>> XQuery/XPath
>> has formal semantics for type-checking using XML Schema types.
>
>
> Is it operational?
>
>> That's
>> *what it does*,
>
>
> *Yes indeed*. (<--I'm not sure about this, but given the extra 
> emphasis, I thought I should reply in kind ;))
>
>> and it does well. It in no way provides a uniform formal
>> semantics for any given XML document.
>
>
> Most surely it does, in the sense that it provides a typing for it 
> (that is, a mapping into a model).
>
>> After all, I can code up FOL or
>> anything else in XML
>
>
> Not in XML or XML Schema alone, I'll bet. Well, I'm not sure about
> the  latter, but imagine you stick to a decidable fragment of XML
> Schema,  then for sure not.
>
>> and the operational semantics of XQuery tell me
>> *nothing* about its operations qua FOL.
>
>
> But this is, well, a lame argument. You're saying that XML's formal 
> semantics isn't expressive enough to formalize FOL. Well, uh, yes. So 
> too for RDF and OWL.
>
>>  However, RDF does provide a
>> uniform formal semantics (although not too interesting!) for any given
>> RDF statement. Now, one can code FOL in RDF, and RDF will tell you
>> things about those FOL statements, but it will be using the RDF model.
>
>
> Sorry, I think you are seriously confused. I totally fail to see that 
> XML is *qualitatively* worse off. It has a formal semantics which
> show  it (well, this isn't clear, but lets assume) insufficient to
> formalize  FOL. So? So too for RDF.
>
> Since various schema languages are expressive enough to encode 3sat,
> I  think you might be surprised at what they can, in fact, say:
>     http://www.idealliance.org/papers/extreme03/html/2003/Lyons01/
> EML2003Lyons01.html
>     http://lists.xml.org/archives/xml-dev/200304/msg00148.html
>
> So, if you coded your propositional logic properly, you'd even get
> the  right semantics!
>
>> What of course we want is to code FOL and get results via the FOL 
>> model,
>> but at least with
>> RDF we have something rather than nothing.
>
>
> We have a something that is probably worse than nothing. The other 
> something might be worse than nothing too, but it's not nothing.
>
> Cheers,
> Bijan.
>
>
Received on Friday, 25 November 2005 19:30:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 16 March 2008 00:11:02 GMT