[
Lists Home |
Date Index |
Thread Index
]
From: "Jonathan Borden" <jonathan@openhealth.org>
> 3) The _syntax_ of OWL has been defined using an _abstract syntax_ which is
> mapped to sets of _triples_. The triples are syntactically defined using the
> RDF N-Triples syntax.
> see the OWL S&AS http://www.w3.org/TR/owl-semantics/mapping.html#4.1
The use of an abstract syntax does not render a language immune from requiring
a standard schema. The purpose of a schema for QA is because a priori we don't
trust that the programmers have done the right thing either with the data or the specs
or the test suites results. Why would an abstract syntax provide any ground for trust?
Schemas can prevent specifications from disappearing up their own Owlholes:
this is the sad condition where something requires so many mapping stages before it
corresponds to bits that answers to basic questions can only be answered by
experts, who will be perplexed that such otiose questions should even be asked.
Examples of basic question might be "Does every owl ontology in XML
contain an element in the OWL namespace, and is this true for OWL Full, OWL DL
and OWL Lite?" Or "Can the RDF elements that are not allowed in OWL DL
appear elsewhere in the OWL document, or is it only under owl: namespace elements?"
> The folks who have written the OWL S&AS are rather intelligent as well as
> experienced and have not fallen into the trap you have discussed above
> precisely given the issues surrounding the RDF/XML syntax and its
> ambiguities (rather there are often several ways to write the same thing
> which makes testing difficult).
I am glad to hear the OWL folks are rather intelligent. The RDF folks who made
an ambiguous grammar in the first spec for RDF were also rather intelligent.
My dog is rather dumb and he also does not make languages with schemas.
> So you see that one person's syntax is often another's semantics...
Sure, but the matter is practical not related to categories: can I take
the serialized form (and XML seems the only game in town here AFA standard
schemas) and validate it usefully using a standard schema language? For basic
RDF structures, OWL is based on the new RDF draft; Dave Becket has done a
really great job on the new RDF and provided a RELAX NG schema. To be
complete RDF should make that RELAX NG schema normative, in particular
to show up what is old and what is new.
Schemas for the OWL elements and the RDF elements that are allowed with
the various OWLs would be useful.
Just as nowadays the onus is on people who don't use XML for data interchange
to justify why they didn't use XML, the onus should be on standards-makers to
jutify why they haven't used a standard schema language. (Actually, standards
makers can do whatever they like: the onus should be on people who adopt
specifications to show why a specification with no standard schema meets
basic QA.)
> For extra credit, I'd like to see a Schematron that can properly
> differentiate between an OWL Lite, OWL DL, and OWL Full document (as defined
> in the above document as an OWL Syntax checker)
Well I certainly don't feel like wasting my time on it! I spent days making
schemas for RDF (I did a DTD and a couple of Schematron schemas)
and decided that the problem is not technical (or IQ related!) but cultural: some parts
of the RDF community are happy to shoot themselves in the foot by having
a Web technology which relentless eschews the tools that help make data
interchange practical and which are appropriate for industrial QA.
Oh, yet another case of NIH: the test documents specify that a syntax-checker returns
the language that has been used, rather than testing whether a document
conforms to one particular language. That excludes the use of out-of-the-box
schema languages; why is that neccessarily?
Syntax is our understanding's friend, not its enemy.
Cheers
Rick Jelliffe
|
