From: Frank van Harmelen ([email protected])
Date: 11/01/01
I think this whole discussion on "levels of compliance/validation" is extremely useful/important, however: I'm a bit surprised about the strong opinions various people have. I am convinced that doing all this stuff on the Web will be very, very different from doing any of this in a more traditional KB setting. (see my slides during the IJCAI panel, on "Semantic Web research challenges to KR"). For example, one of the big attitude changes between old-style hypertext and new-style Web in the early 90's was the treatment of broken links. The fact that the Web doesn't crash on a broken link is a feature, not a bug. It was one of the main reasons that allowed such fast and stable growth. An undefined class is the equivalent of a "broken link" in DAML+OIL. Similar to the move from old-style hypertext to web, I think our attitude to this "problem" will (have to) change when moving from old-style KB to semantic web. I don't claim to know what the best attitude is, simply that I'm very suspicious of using old intuitions/experiences on this new problem. Perhaps defining different levels of "compliance" is the best we can do, and then wait to see how things turn out in the wash. I would be opposed to claiming we already know which compliance levels will be good, required, useful, etc. Frank. ---- (I know, in some of this I'm beginning to sound just like Jim Hendler. I leave it to you to decide if that's a feature or a bug :-)
This archive was generated by hypermail 2.1.4 : 04/02/02 EST