Re: Proposed response to Hugh WInkler - allDisjoint

As tempted as I am to reopen the issue of allDisjoint I won't (as I 
seem to be the main proponent).  However, the result of our LC 
discussion was to add section
   http://www.w3.org/TR/owl-guide/#DisjointClasses
to the Guide.  Any answer to Hugh should direct him to this section 
in our document, not just to a test case.

I would also like to suggest that it is becoming clear that many 
people are clearly going to the reference document without reading 
the guide, I would therefore like to suggest to Guus and Mike a 
positive editorial change might be for the Reference to put a point 
to the Guide link above in section 3.2.4 (DisjointWith) and/or in 
section 5.2.3 (AllDifferentFrom)


Given that this came up so often in our LC comments, and was 
discussed at length by the WG, we should do a better job of making 
sure people can find it (so it doesn't come up again at PR)
  -JH



At 2:20 PM +0300 9/19/03, Jeremy Carroll wrote:
>Hugh was not aware of the solution in test I5.21 002, perhaps this should be a
>FAQ.
>
>http://lists.w3.org/Archives/Public/public-webont-comments/2003Aug/0025
>
>[[
>Dear Hugh Winkler,
>
>the WebOnt Working Group has considered this problem in detail.
>We were aware that a few OWL users will have large numbers of disjoint
>classes, when we last formally considered this (during last call).
>
>As part of the resolution in last call we added the following test to OWL Test
>Cases, which illustrates an O(N) construction equivalent to "owl:allDisjoint"
>
>http://www.w3.org/TR/2003/CR-owl-test-20030818/proposedByIssue#I5.21-002
>
>>  For the 337 terms in our hypothetical UBL Library ontology, we would
>>  enter 56616 <owl:disjointWith> statements. Using the proposed
>>  <owl:allDisjoint> syntax would require 337 statements.
>
>Recalling the formulae from our group discussion:
>http://lists.w3.org/Archives/Public/www-webont-wg/2003Jul/0280
>
>syntax illustrated in test I5.21-002     
>6+6*N
>2028 triples
>
>owl:disjointWith
>N(N+1)/2
>56953 triples
>
>"owl:allDisjoint", like owl:AllDistinct
>2+4*N
>1350 Triples
>
>(In all three, I include the 337 triples needed to declare the classes
>xxx rdf:type owl:Class; in the last I include the triples needed for the
>rdf:List construct).
>
>Thus, while owl:allDisjoint is more efficient, it is by a factor of 50%,
>rather than an order of magnitude.
>
>The Working Group does not intend to make any changes in light of your
>comment.
>
>Please reply indicating whether this is a satisfactory response, copying
>public-webont-comments@w3.org
>
>thanks, and please feel free to make more comments,
>
>Jeremy Carroll
>]]

-- 
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 ***