Proposed response to Alan Rector (QCRs)

Qualified Cardinality Restriction

Dear alan:

   Let me start by thanking you for your comments -- this commented 
generated much discussion in the working group, all available in the 
public archives of www-webont-wg.  I will try to summarize the 
discussion and results here, you are, of course, welcome to explore 
the public archives - some pointers will be provided to make this 
easier.

  Briefly, DAML+OIL had a relatively ad hoc mechanism for representing 
Qualified Cardinalities.  The WG observed that this was a rarely 
used, and hard to describe, feature (not necessarily the notion of 
QCRs, but the odd language features needed to implement them).  We 
thus had decided to remove this feature.

Based on your Last Call comment, we reopened and reconsidered this 
issue.  As you can see from [1], it became clear to the group that we 
needed a better mechanism than what was in DAML+OIL to handle this 
issue (and also to handle some other aspects of qualification, for 
example see [2] which discusses the need to also qualify functional 
restrictions).  A better mechanism than the one in DAML+OIL for 
representing QCRs was proposed by Guus  Schreiber [3], but 
unfortunately it became clear to the group that this would be a major 
change to the language, beyond our current scope.  We therefore 
decided to reopen the issue of Qualified Cardinality Restrictions, 
and then to POSTPONE this issue with a pointer to Guus' proposal. 
The specific decision to postpone is recorded in [4], the rationale 
accepted by the group, and recorded in ourissues document  [5] is 
summarized as:

  	 The Working Group decided 25 Apr 2002 to remove qualified 
cardinality constraints. The issue was reopened due to new 
information Apr 2003 from Alan Rector. In the 8 May 2003 
teleconference, the WG resolved     ... to POSTPONE this issue for 
the following reasons:         * OWL already contains one QCR 
construct: owl:someValuesFrom (QCR with minimal cardinality of 1) 
which covers some frequent-occurring cases of QCRs.         * There 
are some workarounds for QCRs, using the rdfs:subPropertyOf 
construct. These can be used in simple cases, such as the example in 
the Guide below. The WG agrees that these workarounds are more 
problematic for complex part-of relations such as pointed out by Alan 
Rector in his use cases a) and b).         * The evidence on whether 
users need this is mixed. Rector's use cases are compelling, but 
Protege (which has a large user community) has not reported user 
requests for this feature.         * Inclusion of this feature will 
put additional burden on implementations. For example, it is 
nontrivial to add this to Protege.     The Working Group therefore 
POSTPONES the full treatment of QCRs, while considering possibilities 
for making idioms or other guidelines for QCRs available to the 
community.


  We hope that eventual follow on activities to our working group, 
possibly chartered to extend the language beyond DAML+OIL, will 
include a general mechanism for handling qualification.  However, 
adding these to OWL at this time would be a major step, and would 
require significant effort as, in some cases, there is no obvious 
implementation of these properties that can work with the current OWL 
design.

  There are many features that could be added to OWL, and this is an 
important ones, but we've tried to keep the design simple, similar to 
DAML+OIL (by charter) and based on the requirements document that was 
published a few months after we started our group (and updated 
several times, the most recent being [6]).  Thus, the WG did not 
agree to make a major change at this late date.


  We hope you will accept our decision not to add these features at this time.

   -Jim Hendler



[1] http://lists.w3.org/Archives/Public/www-webont-wg/2003Apr/0176.html
[2]
http://lists.w3.org/Archives/Public/www-webont-wg/2003May/0074.html
[3]
http://lists.w3.org/Archives/Public/www-webont-wg/2003May/0072.html
[4]
http://lists.w3.org/Archives/Public/www-webont-wg/2003May/0120.html
[5]
http://www.w3.org/2001/sw/WebOnt/webont-issues.html#I3.2-Qualified-Restrictions
[6]
http://www.w3.org/TR/webont-req/

-- 
Professor James Hendler				  hendler@cs.umd.edu
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  *** 240-277-3388 (Cell)
http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***