Action 236: Review of n-ary datarange proposal

This email addresses my action 236: Review of the following document:

http://www.w3.org/2007/OWL/wiki/Data_Range_Extension:_Linear_Equations

Bernardo

-----

Comments:


- In general, I share Boris's concerns about the implementability of the 
proposal. The references given at the
end of the document are quite general and the relationship between those 
references and the actual proposal is
not clear from the document. Also, the implementation of the OWL 2 
datatypes is already quite complex
as it is. What
additional complexities in terms of implementation effort would this 
extension introduce?

- The actual purpose and scope of the document is also not clear to me. 
The introduction talks about the limitation
of OWL 2 to unary datatypes. The proposal, however, does not seem to 
address the general case of n-ary datatypes.
It seems to focus only on linear equations with rational coefficients 
solved over the reals. Hence,
something should be said about the
relationships and differences between n-ary in general, comparisons in 
general (not necessarily between numbers),
comparisons between numbers, and linear inequations of the sort proposed 
here.

- The kinds of comparisons that can be made is also not clear.
The overview talks about comparisons between the value of one data 
property and the value of another data property.
For the same individual? (e.g. people whose salary is greater than their 
age) For different individuals?
(e.g. women who earn more than their husbands), and so on.

- Section 2 is missing

- The editor note in Section 3.1 is quite cryptic.

- The terms ``Scaled comparison'' and ``linear comparison'' are 
confusing unless properly explained.

- According to the grammar it seems possible to construct a data range 
that, for example, complements a linear comparison.
 In principle this could be easily done by flipping the predicate in the 
comparison, but these kinds of things should
be checked out properly.

- Section 3.2, first sentence. What does ``range refer to''?

- Something should be said about the RDF-based semantics. Does it need 
to be extended? How?

Received on Thursday, 30 October 2008 17:57:01 UTC