[Top] [All Lists]

Re: [ontolog-forum] Requirements of computer language semantics

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Thu, 19 Mar 2009 18:49:25 -0400
Message-id: <49C2CBF5.5090306@xxxxxxxxxxx>
Joel, Ed, and Dick,    (01)

JFS>> The major problems here are not in logic, but ontology...    (02)

JB> I disagree, in spite of the number of issues that are raised.
 > Beyond the "I call this property/you call this an attribute"
 > discussions, there is a common understand of "measure", and yet
 > I have not found a consistent and strong way of representing it.    (03)

That is indeed ontology.  We all have a vague notion of 'measure'
along the lines of what you'll find in a common dictionary.  But
when you get to all the detailed issues, the ordinary notion leaves
the most critical issues unspecified.    (04)

JB> So just like "2.0"^^xsd:float maps the character string "2.0"
 > into the number 2.0, "si:meter" not only maps it to same numeric
 > space as xsd:float but also to si:length as the UnitsML quantity
 > "length" (which doesn't have a URI that I've been able to find).    (05)

Those are all issues of ontology.    (06)

EB> OK.  If you are mapping a declarative knowledge representation
 > language to an axiomatic first-order language, then I agree.
 > Those are strong constraints on L1 and L2 that won't generalize
 > much.  For example, they don't generalize to L1 condition-action
 > languages, even though most of the actions may be:  conclude <S>.    (07)

I'm happy that we agree on those issues, because I was only talking
about declarative languages of the sort defined by the Semantic Web
and the Common Logic standard.    (08)

EB> Into CL, yes.  Into Ln (unspecified) is quite another thing
 > entirely.    (09)

For the present, I would be quite happy to establish interoperability
among the languages that can be translated into CL.  Nobody has ever
succeeded in developing anything remotely adequate for procedural
languages -- including the condition-action languages and the
procedural parts of languages like Prolog.    (010)

One of the major problems that limited the usefulness of Corba is
that they got stuck in the procedural swamp.  They had to treat
*every* release of *every* brand of C++, for example, as a distinct
language.  I would never dream of trying to reconcile that mess
of languages.    (011)

What I would recommend for interoperability among systems
implemented in procedural languages is to agree on declarative
specifications of the *message types* that can be interchanged
among those systems.  That is what SOAP does, and it's far easier
than getting into the internals of the languages, as CORBA did.    (012)

EB> The problem as stated was to define the semantics of a given
 > language L1.  ISO 24707 does exactly that, where L1=CLIF, using
 > the Tarski approach without reference to any other language.
 > And the RDF and OWL Recommendations do something very similar.    (013)

Yes.  But after you have a definition of a general language like
CL, then you can define a large number of other languages, including
RDF, OWL, and many other by a translation to CL.    (014)

JFS>> I would not recommend the alternative of trying to prove that
 >> n independently defined model theories are compatible.    (015)

EB> Nor would I.  Nor do I see how it is relevant to the issue of
 > providing formal semantics for a given language.    (016)

As I said earlier, if you are only going to use language L in
isolation from all others, then you can define a model theory
for L without any concern for interoperability.    (017)

But if you want to define two or more languages that must
interoperate with one another, then it is better to define
their semantics by translation into some common superset.    (018)

And by the way, that approach can also support interoperability
among procedural languages, as GNU and MSFT do with their common
runtime environments for procedural languages -- but they only
guarantee interoperability for their own dialects.    (019)

RHM> The reason I cannot be more specific is that the details
 > depend upon the purpose of the person who is creating the
 > propositions...    (020)

That is a cop out.  The rest of us have been defining languages
that enable astrophysicists, chip designers, historians, ...,
to state exactly what they mean by their intended context.    (021)

John    (022)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (023)

<Prev in Thread] Current Thread [Next in Thread>