Re: WebOnt Plans (requirements: yes, let's)

"Peter F. Patel-Schneider" wrote:
> 2/ What are the high-level requirements for the final ontology language.
> Again we cannot come up with a final answer now, and probably won't be able
> to until we are done, but I think that having some high-level requirements
> now would make our work much easier.

I second this suggestion too.

[...]
> [Note that I just made most of these up off the top of my head.  Some of
> them do not have my support.]

I'll weigh in, if only to provoke a bit of discussion...


> The kind of requirements I am think of would be something like:
> 
> a) There will be an XML surface syntax for the language.

I'm willing to accept that one without further discussion.

> b) All RDF/XML documents will be accepted by our language.

I don't know what that means, but I think it would
be useful to discuss it.

> c) The surface syntax parsing will be compatible with XML parsing and
>    validation.  It will be possible to use XML infosets or XML data models
>    with post-schema validation information incorporated that have been
>    generated by unmodified XML and XML-related parsers and validators as
>    inputs to our post-syntactic processing.

Ugh... I hope we keep a safe distance from the
post-schema validation stuff.
It's really quite complex, and the WG itself hasn't figured
out how to formalize it yet.
Not to mention the trust/availability/mapping issues...
note that the connection from XML documents to schemas
is many-to-many. I hope we preserve the critical
feature of RDF documents that their meaning
as a logical formula is self-evident.


> c) There will be model theory for the language.

I've been convinced over the last 6 to 18 months that
this is worthwhile, if not critical.

I find axiomatic semantics easier to deal with (because
I get more help from the machine) but I can see
the value of having both.

> d) The model theory will be compatible with the RDF model theory, to the
>    greatest extent possible.

That's my hope. I'm just starting to understand what it means, though.


> e) If there is a conflict between RDF/XML syntax compatibility and RDF
>    model theory compatibility, then compatibility with RDF model theory
>    will have precedence.

I doubt such a rule would be useful; that is: I doubt
it would help folks to buy into the resulting design.


> f) The language will have decidable entailment.

I don't have any need for decidability; I really hope
the discussion of use cases and such will help me
(a) understand why folks find this so valuable,
and (b) convince them that it's not necessary ;->






-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Wednesday, 28 November 2001 17:05:05 UTC