Comments Regarding Built-ins and OWL-RIF

I am currently working with OWL/SWRL and other rule systems and I am
glad to see the high level of technical specification in the new July 3
draft RIF proposals.

 

Regarding the "Datatypes and Built-Ins" document
(http://www.w3.org/TR/rif-dtb/ <http://www.w3.org/TR/rif-dtb/> ), I am
interested in the differences between the list of RIF built-ins versus
the older SWRL submission (http://www.w3.org/Submission/SWRL/
<http://www.w3.org/Submission/SWRL/> , 2004).  I see that a lot of the
same primitives are available under RIF as under SWRL. 

 

I am glad to see that the selection of list-manipulation primitives is
expanded over the SWRL set, since I am finding that in many rule
applications, the construction of complex result sets in rules is much
easier using lists. 

 

For the construction of more complex results, a question has been raised
on some forums about why SWRL does not have an "individual create"
built-in to allow the construction of individuals in the head of rules.
This also could have an application in data integration cases where the
construction of the target data has a substantially different structure
than the source data. Has this issue been raised in the RIF committee?

 

I also notice that some of the SWRL numeric calculation primitives have
been pruned from the list in RIF (trigonometric and most of the
round-off functions), leaving mostly basic arithmetic operations in the
list.  Is the intent of this to force complex calculations outside the
rule engine (where they can be done more efficiently)?

 

Regarding the http://www.w3.org/TR/rif-rdf-owl/
<http://www.w3.org/TR/rif-rdf-owl/>  document, I have found that if a
standard does not specify enough concrete examples, it leads to
confusion when developers try to use it. Aside from the concrete
examples in the SWRL document (the RDF/XML and OWL/XML samples), there
was a lot of confusion on the forums even as late as this year about
basic syntax usages for various aspects of SWRL rules. (I found it
necessary to start developing and sharing a set of working samples in
various dialects to get around this problem, but it is very late in the
game to be doing this.) Given that there are concrete XML examples of
RIF in various flavours, and the rif-rdf-owl document specifies the
limits of how the documents can be "combined", is that enough?

 

Many thanks to the Working Group members for their excellent work on
these documents.

 

Alan Meech, Senior Consultant, CGI (Canada)