Re: Feature synopsis: content comments (named cardinalities)

On Wed, 2002-09-04 at 03:45, Jerome Euzenat wrote:
[...]
> 2.1.4: OWL Lite cardinality
> 
> 	I think that this depends on the policy:
> - having a minimal set of primitives (but then why cardinality and 
> FunctionalProperty)
> - providing no-brainer restrictions for the user of OWL-Lite.
> 
> 	I take the second standpoint: It is very difficult to refrain 
> fram using (cardinality 3). If you look at the combonatorics of these 
> primitives, we can provide 3 values for each min and max cardinality 
> (NOVAL, O and 1). This generates 9 possiblities (and I do not count 
> cardinality).
> Of these only 4 are useful:
> - (maxcard 0 || maxcard 0 && mincard 0)
> 	[0 Ø] EMPTY // alg. nothing
> - (maxcard 1 && mincard 1)
> 	[1 1] SINGLE(TON) // alg. X
> - (maxcard 1 || mincard 0 && maxcard 1)
> 	[0 1] FUNCTIONAL // alg. X?
> - (mincard 1)
> 	[1 +inf] NONEMPTY // alg. X+
> - (mincard 0 || VOID)
> 	[0 +inf] NOINFO [this is useless] // alg. X*
> - (mincar 1 && maxcard 0)
> 	INC [this is an error] // alg. does not exist
> 
> Retaining, empty, single, functional, nonempty as four primitives 
> replacing maxcard, mincard and card would strongly reduce the risk of 
> errors from the user standpoint:
> - by avoiding cardinality 4
> - by avoiding the use of multiple expressions (meaningful: mincard 0 
> && maxcard 1 or meaningless: maxcard 0 && mincard 1).
> - more generally by requiring at most one keyword for expressing any 
> meaningful property!
> It would not put additional burden on the implementers.
> Only the documentation developers will have a bit more work.
> 
> I would be greatly in favour of this solution.

I tend to agree; in fact, thanks for the prod; I ought
to raise this as an issue.

The WG discussed this in July in Stanford;
at the time, a proposal for named cardinalities,
rather than min/max, wasn't very well developed, so we fell
back to the DAML+OIL status quo:

"""
DanC [or was it Deb?]: PROPOSED: hasExactlyOne, as in BaseballTeam
hasExactlyOne pitcher. likewise for hasAtMostOne/hasAtLeastOne

IanH: do we really want to add this sort of new, redundant terminology?

MikeS: indeed, let's stick to owl:cardinality, restricted to 1 or
owl-lite

straw poll: 8 to 5ish

3ish would object to cardinality "1"

RESOLVED: to go with owl:cardinality/min/max, restricted to "1",
dissenting: Evan. abstaining JimH abstained, Deb, Jos, Connolly.
"""


--  http://www.w3.org/2002/07/01-webont-irc
 http://www.w3.org/2001/sw/WebOnt/ftf3.html


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Wednesday, 4 September 2002 12:47:31 UTC