RE: Imports issue

I am indeed in the process of adding OWL support for Jena2, and we are
(were? :-) planning to add better support for imports processing.  My plan
is to have move the imports closure processing from the existing position
inside the DAMLModel to a new ontology document manager class.   This would
provide a means for caching local copies of documents from remote URL's, and
a place to specify a policy for processing imports.  The default policy
options will be "do" and "don't" :-), but once the framework is established
we can add more sophisticated control in future.  Using this approach, it
would also be possible for users to pre-compute and cache the imports
closure, making ontology loading (hopefully) faster.   Oh, and yes the
current DAML loader does an occurs check for import cycles, and the new code
will do so as well.

I'm concerned though that I may have missed some subtlety in OWL imports.
It doesn't look impossible to me - so what am I missing?

Cheers,
Ian


> -----Original Message-----
> From: Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com] 
> Sent: 20 February 2003 10:02
> To: Peter F. Patel-Schneider
> Cc: schreiber@swi.psy.uva.nl; www-webont-wg@w3.org; 
> ijd@hplb.hpl.hp.com
> Subject: Re: Imports issue
> 
> 
> I agree with Peter.
> 
> Jena  has implemented daml:imports for ages now (well over a 
> year). It is limited to what fits in memory (with Jena's 
> somewhat profligate 
> approach to memory).
> 
> I am unsure as to whether it implements the transitive 
> closure as specified 
>   for OWL - i.e. if it is robust against import cycles.
> I am copying Ian Dickenson, the developer of that code: Ian 
> please can you 
> comment (to the webont list) on how far or close it is from 
> the OWL imports 
> mechanism.
> Peter outlines the OWL algorithm below.
> 
> Ian is working on OWL support for Jena2, and I understand 
> that we intend to 
> continue to support imports.
> 
> As an aside - I think imports is a much more important 
> feature in OWL Lite 
> and OWL DL than in OWL Full.
> 
> Jeremy
> 
> 
> 
> Peter F. Patel-Schneider wrote:
> 
> > From: Guus Schreiber <schreiber@swi.psy.uva.nl>
> > Subject: IImports issue
> > Date: Thu, 20 Feb 2003 00:26:57 +0100
> > 
> > 
> >>I'm feeling increasingly uncomfortable about our "imports" 
> resolution
> >>(see the discussion threads cited in the agenda).
> >>
> >>Unless we get in the very near future clear evidence this is an
> >>implementable language feature, I will have to reopen this 
> issue and 
> >>propose to give imports the same informnal status as the 
> versioning stuff.
> >>
> > 
> > Huh?
> > 
> > To implement imports, it suffices to modify an RDF/XML processor as
> > follows:
> > 
> >    Whenever an imports triple is found, first check to the 
> if the object of
> >    the triple has been imported already.   If not, get the 
> document that is
> >    pointed to by the object of the triple and run it 
> through the RDF/XML
> >    processor.  Then merge the result with the current 
> graph.  Only a very
> >    small amount of care is required to prevent loops.
> > 
> > What could be easier?
> > 
> > What is currently being argued about is how imports 
> interacts with OWL 
> > Lite and OWL DL, i.e., what documents containing imports 
> count as an 
> > OWL Lite or OWL DL document.
> > 
> > 
> >>Note that responses of the type "this is a useful/necessary 
> feature" 
> >>are
> >>not helpful at this point. 
> >>
> > 
> >>If we cannot show that imports  can be
> >>implemented, we will not be able to go to Proposed Rec with 
> the OWL spec 
> >>as it stands. 
> >>
> > 
> > This is crazy!  Who has claimed that imports is not implementable!
> > 
> > 
> >>We would all hate toi see that happen.
> >>
> >>Guus
> >>
> > 
> > peter
> > 
> > 
> 
>

Received on Thursday, 20 February 2003 05:43:00 UTC