QCRs - How much time are they worth?

WOWGers-

Summary:  Adding QCRs to OWL is likely to add 3-4 MONTHS to the time 
until we move OWL to PR.  Are they worth it?  What is an alternative?

Details:

It is clear that every proposal to date for QCRs would require more 
than syntactic sugar.  Due to the nature of the graphs we have chosen 
to use to date, we would either need to add new constructors (like 
the Q constructors in D+O), make changes to our treatment of current 
constructs (like changing all restrictions to be qualified in some 
sense), or invent some new syntax that requires reworking of 
normative sections of our documents (such as Ian's most recent 
proposal).  While none of these require major semantic change, they 
all require changes at the syntactic and graph level that go beyond 
what could comfortably be defended as "editorial" change.

We then add to this fact that, although there have been 
implementations of QCRs, none have been doen in the correct syntax 
(as we don't yet have one).  Worse, going over the list of D+O tools 
that I had hoped to cite in our implementation report, it is unclear 
any of them actually handle the Q-constructs in any real sense.  The 
validator allows them, but doesn't check for violations, the 
"instance" based tools appear to check cardinality constraints in 
some form (for example RIC enforces cardinality, but not qualified 
cardinality), and their is no evidence in the toolset on the D+O page 
that a reasoner that handles the qualified cardinality constraints 
(including instances) exists.

As such, we would almost definitely need to do the following:
  Upon end of current LC, we would need to make changes for QCRs and 
do a second LC  --  If nothing else tests for QCRs would have to be 
added to the Test document.  It would be a mistake to release Test as 
LC now if we know it is going to change with a new feature that 
requires new tests and changes old ones.
  This means at end of current LC we add a second LC period of a 
minimum of 4 weeks.
  On close of that LC, we would have to move to CR (not PR).  While we 
might be able to get away with a 4 week CR period, a longer CR period 
is typical, and there would be time between the second LC and the new 
CR.
  Thus, I believe our schedule would need to be

May 9, close current LC.
June 1, new LC begins
end of June - new LC ends
mid-July (earliest) - CR begins
end-August - CR ends   (note, this is optimistic as getting much done 
in Aug is hard)

This means rather than having OWL at Rec by end of summer, we'd be 
looking at having it at Proposed Recommendation by September.


Alternative:

   Since OWL DL is SHIQ (maybe SHIOQ) we know QCRs are in our current 
semantic coverage.  This means that we should be able to develop an 
idiom within our current language to handle them (even if it means 
creating extra classes).  We have talked often about the need for an 
idioms page.  We could create that page, and make a recommendation 
for how to handle QCRs be one of the entries on that page.  We could 
have Guide, Ref and Features (non-normative documents) point at that 
page and have discussion of a recommendation for QCRs.
  Note that this is consistent with past practice of our WG, within 
W3C process, and would not add any delay to OWL release.

Proposal:
  I propose we change the closing status of issue 3.2 to POSTPONED. 
We create the idioms page and a discussion of QCRs go there.  A 
pointer to that discussion is added to the issues page, and can be 
referred to in our non-normative documents if desired.




-- 
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	  240-731-3822 (Cell)
http://www.cs.umd.edu/users/hendler

Received on Monday, 28 April 2003 08:16:36 UTC