Re: Comment on RDF Mapping: variables in sequence pattern

[Again, not an official reply.]

It does appear to me that the rules for reverse mapping of sequences are
not complete.    Two lists that share an intermediate node can be
(non-deterministically) accepted.  This includes strange lists that loop
back to themselves (and that also have a valid tail).

A "big club" fix would be to have a global constraint along the lines
of:

	No blank node can be used in more than once in these patterns.
	(This means that all lists are non-cyclic and do not share
	tails.)

It might also be possible to have a more targetted fix, along the lines
of the method used when parsing class expressions.

peter




From: "Mike Smith" <msmith@clarkparsia.com>
Subject: Comment on RDF Mapping: variables in sequence pattern
Date: Tue, 6 Jan 2009 09:53:10 -0500

> 
> While verifying some WebOnt test cases against the RDF to structural
> mapping defined at [1], I noticed that there is no constraint
> preventing variables within the sequence pattern from matching the
> same node (see the second row of Table 3 at [1]).  I found this
> problematic, particularly when trying to avoid things like cyclic
> lists (as in the nonconclusion ontology of [I5.5-006]).
> 
> I believe that adding the constraint to the mapping document will
> clarify the expected behavior.
> 
> -- 
> Mike Smith
> 
> Clark & Parsia
> http://clarkparsia.com/
> 
> [1] http://www.w3.org/TR/2008/WD-owl2-mapping-to-rdf-20081202/#Mapping_from_RDF_Graphs_to_the_Structural_Specification
> [I5.5-006] http://km.aifb.uni-karlsruhe.de/projects/owltests/index.php/TestCase:WebOnt-I5.5-006
> 
>

Received on Wednesday, 7 January 2009 18:23:02 UTC