Issue-144: missing base triple in serialization of axioms with annotations

We had quite a long discussion about this during last week's teleconf  
[1], and I was hoping for some followup via email. To kick things  
off, I have attempted to summarise the discussion below. No doubt the  
protagonists will correct me if I (unintentionally) misrepresented  
what they were saying.


Zhe: Adding triple will ease burden on implementers

Boris: May be mapping issues; what to do with an RDF doc that  
includes a reified triple without the corresponding base triple? Not  
likely to occur often in practice, so not much burden.

Michael Schneider: More stable without added triples; may not be  
possible to add to ontologies where we don't have write permission;  
might cause problems with axiom closure; not in favour.

Alan: doesn't see a problem with missing base triple, or with issues  
raised by Michael; issue is with monotonicity rather than performance.

Boris: OWL RL reasoner would miss inferences without base triple; no  
guarantee that it will be included; regarding monotonicity, OWL Full  
semantics entail base triple from reified one.

Peter: Alan's example not non-monotonic.

Zhe: still wants to stress performance; doesn't want to check  
information on all triples; what if there is a mixture of annotated  
and not -- do we forget the ones without annotations.

Mike Smith: Such axioms are structurally different; spec already  
deals with this.

Boris: Possible solution is to suggest "physical proximity" of  
reified triples and have parser add base triple.

Alan: RDF pipes could be a problem.

Boris: Doesn't see this.

Peter: On performance, I/O cost of larger files may outweigh  
increased cost of de-reification rule.


Received on Monday, 15 September 2008 11:39:27 UTC