Hi Pat, On 12 Jun 2011, at 00:33, Pat Hayes wrote: >>> Nothing is gained from the range assertions. They should be dropped. >> >> They capture a part of the schema.org documentation: the “expected type” of >> each property. That part of the documentation would be lost. Conversely, >> nothing is gained by dropping them. > > Let me respectfully disagree. Range assertions (in RDFS or OWL) do *not* > capture the notion of "expected type". They state a strict actual type, and > cannot be consistently be "over-ridden" by some other information. Which has > the consequence that these are liable to be, quite often, plain flat wrong. > Which in turn has the consequence that there is something to be gained by > dropping them, to wit, internal consistency. They are not mere documentation; > they have strictly entailed consequences which many actual reasoners can and > will generate, and which to deny would be to violate the RDFS specs. If you > don't want these conclusions to be generated, don't make the assertions that > would sanction them.
Data on the Web is messy. You cannot reason over it without filtering it first. I think it is useful to document how data publishers are *expected* to use these terms, even if we know that many will -- for good or bad reasons -- use them in *unexpected* ways. > For documentation, use the structures provided in RDFS for documentation, > such as rdfs:comment. rdfs:comment is for prose. We explicitly know the “expected types” of properties, and I'd like to keep that information in a structured form rather than burying it in prose. As far as I can see, rdfs:range is the closest available term in W3C's data modeling toolkit, and it *is* correct as long as data publishers use the terms with the “expected type.” Best, Richard