Re: AW: [ontolog-forum] Current Semantic Web Layer Cake

Kuldar and Valentin,

I wholeheartedly agree:

 > It is not possible to solve the interoperability problem
 > by imposing a set of standards "from above".  The world is
 > too diverse a place to assume that everyone is going to
 > use them.

The best standards are the ones that develop "from below"
through widespread use.  Then the job of the standards
committees is to recognize what is already being used,
to tidy up the details, and to solidify the foundations.

That is why I was so appalled by the idea of proposing a
standard for syntax with the semantics left up for grabs.
We should have well-defined standards for systems that have
proven their value in practical applications and widely
available implementations.

For logic, there are two extremely important de facto
standards:

  1. Classical first-order logic, which has many widely used
     subsets and supersets:  Aristotle's syllogisms, which
     support the ubiquitous type hierarchies; description
     logics, which are supersets of syllogisms that include
     OWL and other popular systems; a large number of rule-
     based systems; specification languages, such as Z and
     the Object-Constraint Language (OCL) of UML; and many
     other systems that are used for ontology and AI.

  2. A variant of first-order logic whose semantics is defined
     by "negation as failure".  The most widely used of all such
     systems is the SQL query and constraint language, which
     drives the databases that run the world's economy.  Prolog
     is a superset of SQL, which is used in large-scale commercial
     and government applications.  Prolog is also an ISO standard,
     and its semantics is a superset of many rule-based languages
     other than those in point #1 above.

The theoretical foundations for these two versions of logic
were established many decades ago, there are ISO standards for
both of them, and there have been widely used implementations
and applications for over 40 years for classical FOL and over
30 years for SQL and Prolog.

There are also many other systems whose semantics is not well
defined.  But the overwhelming majority of well-defined logic
systems fall into one or the other of the two cases above.
Therefore, these are the candidates that are ready for
standardization today.

Defining standards based on these two does not mean that
others cannot be added later.  But they would have to wait
until there is some consensus on their semantics, and they
have been successful in widely used practical applications.

For further discussion of these and other issues, see

    http://www.jfsowa.com/pubs/fflogic.htm
    Fads and Fallacies about Logic

John Sowa

Received on Saturday, 4 August 2007 03:17:07 UTC