Re: Response to QA comments, Comment on QA draft

Jeremy Carroll
> >>   Guideline 9. Allow extensions or NOT! [6]
> 
> Perhaps I don't understand ...
> 
> I have been thinking for a bit that OWL DL is not perfect but ...
> 
> e.g. QCRs well, one could have a nonstandard extension to OWL DL that 
> supported them.
> 
> e.g. cycles of bnodes forming unnamed individuals; one could permit them in 
> a nonstandard extension
> 
> e.g. predicates involving more than one data value; one could permit them 
> in a nonstandard extension
> 
> e.g. URIs for user defined XML Schema simple types; Jena already does allow 
> them in the de facto non standard extension
> 
> Obviously one needs a strict mode which switches any of this off, but I am 
> not sure what simply forbidding extensions buys us.
> 
> It seems more to the point to say that extending OWL DL, takes you into OWL 
> Full, in which the degree of interoperability expected is lower. Such 
> extensions seem to be natural.

Perhaps we need to distinguish between syntactic and semantic
extensibility.  

Any extension to the syntax of Lite takes you into DL, Full, or Other;
any extensions to the syntax of DL takes you to Full or Other; any
extension to the syntax of Full takes you into Other.  "Other" is not
interoperable, is not OWL, is not RDF/XML, and is not in conformance
with the OWL Specs.

Meanwhile, Jim says OWL Full is extensible, and in some sense I agree,
but I think that's semantic extension by assigning certain URIs
meaning, much like OWL Full extends RDF.  In that sense, Lite and DL
are also extensible as long as you avoid the forbidden URIs that would
take you into another species.  If I define some URIs for talking
about books, I've extended the language to allow talking about
books....  but that extension is completely orthogonal to OWL, so it
doesn't seem like an extension at all.  It's more like the OWL
vocabulary and the Bookstore vocabulary can both be used together in
RDF.  Even if the additional vocabulary provides a logic language
which is designed to nicely co-exist with OWL, it's still just another
vocabulary being used along side OWL.

This might be an enormous rathole.  If there is disagreement about
this, let's not get sucked in purely on account of extensibility being
one of the QA guidelines.

     -- sandro 

[6] http://www.w3.org/TR/qaframe-spec/#Ck-extensions-disallowed

Received on Thursday, 12 June 2003 09:07:36 UTC