Re: OWL reasoner, validator, editor ...

On Mon, 2003-10-06 at 17:34, Bernard Vatant wrote:
> Follow-up to my "awkward" croosposting from Protégé discussion list, and
> previous private answer to Dan.
> > crossposting is awkward.
> Apologies again for the mess.

No problem; It wasn't that big of a mess.

>  The current message is only to this WG.
> > And it's not clear that this is WG business.
> Well, was not completely clear to me when I posted it, but I'm not sure it
> is not this WG task to define some guidelines about various tools that are
> likely to be used to deal with OWL in various ways, to avoid people the
> kind of confusion I've myself been into for a while (and maybe still am).
> > Does Protege claim to export only consistent ontologies?
> No, of course currently it is not. But users would and will expect from a
> software calling itself an ontology editor, providing an OWL export
> facility, some kind of control at editing time. And at least some hints on
> what is checked and/or what is not.
> > > OWL validators as well, such as
> >
> > I don't believe that tool claims any sort of completeness.
> No, it is not. But what is misleading is that users see the tool detect
> some types of inconsistencies, so will improperly and joyfully infer that
> it detects *all* inconsistencies.
> > Perhaps "validator" is a misleading term. It's not defined
> > in the specs. The spec uses terms like "complete OWL DL
> > consistency checker".
> Agreed. So what should we do about those misleading terms? Is it this part
> of this WG business to provide recommendations about the minimal level of
> semantic control an application has to provide to be called an "OWL editor"
> or an "OWL validator" or other qualificatives that are likely to pop up
> ("OWL checker", "OWL wizard" ...).

It has been the business of the WG on occasion...

"The WG discussed the costs and benefits of specifying conformance of
classes of software in addition to the (unanimously supported)
specification of classes of documents."
  -- Minutes Jan 2003 ftf

To reiterate my position from that discussion:
my experience is that far too often, document formats are specified
in terms of what various bits of software do with them, leading
to piles of unanswered questions about what other sorts of software
could or should do with them. So it's vitally important to specify
what the documents mean largely, if not completely, independent
of what various bits of software does with them.

Our charter is pretty clear on our responsibility to do that:

  The products of this working group must be supported by a formal
  semantics allowing language designers, tool builders, and other 
  "experts" to be able to precisely understand the meaning and
  "legal" inferences for expressions in the language.

Software conformance clauses are not ruled out by our charter,
but they're not required either.

>  Or shall we look away from those
> disgusting things, stick to "complete OWL DL consistency checker", and let
> implementers and users sort out the confusion that will result from all
> those applications providing only partial and ill-defined semantic control?

That's my preference. I don't have any reason to believe I can do
it better than they can.

> On Protégé list, Peter Crowther answered : "Given the many, many more
> complex cases of inconsistency, it doesn't seem sensible to give error
> messages for the trivial ones and not for others." I'm not sure I agree
> with that from the  mainstream user viewpoint. People are used to several
> levels of checking, like XML well-formed vs XML valid against a scheme,
> orthographic checkup vs grammatical checkup, etc ... Does not it make sense
> to envision OWL editors checking at editing time for trivial local
> inconsistencies, and defering to more powerful reasoners the task of
> detecting the more difficult inconsistencies?

Well, no, it does not make sense, to me. That is: I do not know what
you mean by "trivial, local inconsistencies."

Dan Connolly, W3C

Received on Monday, 6 October 2003 20:37:14 UTC