Triples and OWL2

Hi,

I'm not a regular member of this forum, but as a developer of the
SWI-Prolog Semantic Web infrastructure I hear some things ... some
of which are worrying ...

I have had some discussions about OLW2. Many of the semantic extensions
have my support, although I doubt the number of users that actually
needs them is sufficient to warrant early standardisation at this level.

What is worrying me is to destroy the semantic web stack, going

  (1)	serialisation --> triples --> semantics

The OWL2 initiative seems to change this entire stack, producing
something like this (with some complicated scenario that guarantees some
level of compatibility):

  (2)	serialisation --> triples --> semantics
			                 ^
			  another serialisation

[if you use a variable-pitch font: <another serialisation> points
*directly* to <semantics>]

This upsets my entire toolchain, which is based on the assumption of
model (1). Now, of course, some people may think this is just my
problem. I believe this is not the case. If we re-introduce the
immediate mapping from syntax to semantics, we are likely to fall into
the same trap from which knowledge representation formats suffered for
decades: different syntaxes with 90% semantic overlap that are hard to
unify in a single application (I assume OWL2 is not the final KR
language).

The power of the semantic web is a stack with a single mandatory
serialisation to a simple -but powerful- data-model. On top of that
model we created a stack of languages with increasing semantic
expressivity using the push-down mechanism.

The semantic extensions do not seem to justify breaking this stack.
I think the task of OWL2 is to provide an extension to the semantic
interpretation of the triple model of OWL1, possibly fixing some OWL1
mistakes at the same time.  This working group should NOT be about the
serialisation.

	Regards --- Jan

Received on Friday, 30 January 2009 20:31:40 UTC