- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Mon, 5 May 2003 18:37:57 +0200
- To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
- Cc: <www-webont-wg@w3.org>
> > > > Comments? > > The reason that I'm opposed to the change is that it would mean that these > nodes would not have typing information associated with them. The > requirement for typing information for all nodes in OWL DL came > from you, I > believe. If you think that typing information is not required for list > nodes, then why did you agitate for this requirement? > I still think that it is better to have the typing information on the list nodes, but this is primarily an aesthetic judgement, and an ease-of-use judgement and it is not a judgement about *needs*. My view remains that in OWL DL there appears to be a *need* for a large number of nodes to be explicitly typed. There are some nodes that do not *need* to be. For user simplicity I think the current explanation is best: all nodes, other than built-in URI refs and literals are explicitly typed. Each exception to this, is possible, but at some cost of user confusion (in my view). The List triples, since they are usually hidden from the user, are perhaps less confusing than some other possible exceptions. So my preference would be to retain the current syntax rules for rdf:parseType="Collection" and retain the requirement that all nodes be explicitly typed. However, it is a preference - the real requirement is a clear articulation of which nodes need explicit typing, ideally with some sort of rationale. (I find rationales for *all* easier than rationales for 80%). I can see that from RDF or OWL Full, these triples just appear as redundant junk - my take is that this syntax is there principally for OWL DL and OWL Lite where I do see advantages in terms of uniformity. However, this is an aesthetic judgement, and I am not the arbiter of taste. I also point out, that we have, in the past, found surprising examples which have demonstrated a need for type triples in other places. Caution would argue for retaining the type triples on lists - but I don't really believe there is a need. Jeremy
Received on Monday, 5 May 2003 12:38:24 UTC