Re: do we really need two languages/levels? [Issue 5.2]

On July 11, Raphael Volz writes:
> I can only second Chris' statement
> "If you use a feature, all the time or not, that is not in OWL-lite, then
> use heavy OWL.  "Removing" a feature from OWL-lite is not removing it from
> OWL."
> However, the argument for having two conformance layers is not only
> restricted
> to "cheap admission".

We discussed this at length in Toulouse, and concluded that the "cheap
admission" FOR IMPLEMENTORS is the only argument that really stands

> In many cases conceptualizations will simply not require the given
> expressive power,
> consider for example WordNet or any other large thesaurus that have found
> broad user
> communities and may be called ontologies (since they establish shared
> agreement due
> to common usage). If we know apriori that only a limited subset of language
> features
> is used, different (considerably faster) evaluation strategies can be used
> in
> implementations.

But rather than having pre-defined conformance levels, wouldn't it be
better simply to use a reasoner that was appropriate to the set of
language constructors you happen to be using? In that way, many
different reasoning systems/paradigms might be used. Even better would
be to use the range of language features to automatically select from
the range of reasoning services that are on offer - i.e., providers of
reasoning services describe the capabilities of their services,
consumers of reasoning services describe their requirements, and
appropriate services are discovered. Sounds strangely familiar doesn't
it :-).

> Third, the effort to learn the language is tremendoulsy simplified. Having a
> lower
> barrier for membership of the expected community will certainly increase the
> size
> of the community.

Again, rather than having predefined conformance levels, what is wrong
with simply having members of various user communities produce their
own guide to the subset/usage of the language that is appropriate to
them? This is much more flexible, and would serve the purpose just as
well if not better.


> Raphael

Received on Thursday, 18 July 2002 08:55:17 UTC