- From: Dave Beckett <dave.beckett@bristol.ac.uk>
- Date: Thu, 18 Jan 2001 20:38:46 +0000
- To: uche.ogbuji@fourthought.com
- CC: Dan Brickley <Daniel.Brickley@bristol.ac.uk>, Dennis van der Laan <LaanD@vertis.nl>, www-rdf-interest@w3.org
>>>uche.ogbuji@fourthought.com said: > It seems that ever since Java came out crowing for some odd reason > that it had "fixed" C++'s use of MI, the orthodoxy has become that > multiple inheritance is bad, but I think this is nonsense. There are whole newsgroups where you can have fun with that flame war. > Certainly in RDF it makes the most eminent sense. Inheritance is a simple > arc, and I don't see how the cardinality of that particular arc hampers the > quality of the expression or implementation in any way. In off list discussions, Dennis and I emailed further about my comment and I would like to restate it in this way - RDF has a class and type system that is neutral on how the class graph is arranged. There is only one restriction - no loops of subClasses are allowed. i.e. hierachies, trees including multiple inheritance are allowed. Neither of the standards really mention inheritance at all - RDF M&S mentions it once in the introduction without much indication of what that means. As an implementer, I just need to know - what typing / classing is allowed and then I can try to work out how best to optimise it. In this case, since there are no loops, a partial ordering is one possible method. I've reworded some of my original comments in http://www.redland.opensource.ac.uk/notes/concepts.html and added more explanations. Dave
Received on Thursday, 18 January 2001 15:39:02 UTC