- From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
- Date: Wed, 24 Oct 2001 08:37:58 -0400
- To: w3c-rdfcore-wg@w3.org
[This note is stated in rather blunt language. I am doing this on purpose, as I feel that it may be the only way for the RDF Core WG to see what is going wrong here.] I have a whole ton of comments on the reification section, but first I'll get straight to the point: It's terrible. Why? Because it contains many of the same old fallacies about RDF reification that have been spouted ever since RDF came into being. Let's consider some of the fallacies in the section: - In (p2) the section states that reification lets us ``make a statement the subject of another statement''. It does no such thing. At best, it lets us make a resource that has some relationship to a statment the subject of another statement. This fallacy is repeated in (p22). - In (p7) the section indicates that the reification of a statement is supposed to denote a statement. However, there is no denotation relationship between the reification of a statement and the statement itself. (Further, denotation is a formal relationship between a piece of syntax and a semantic entity. The reification of a statement is not a piece of syntax, so it doesn't denote anything.) - The subject, predicate, and object properties of a resource are just properties. They don't have any connection to any (pre-existing) statement (p8). - The section talks about THE reification of a statement (p10), but there is no restriction that I can see that requires that there be only one reification of a statment. - Statement reification does not help RDF to be self-describing (p14). - There is no distinction between having "eric" as the object of <rdfprimer> <editor> "eric". and as the object of _:stmt <object> "eric". Both use "eric" in exactly the same way. Thus the distinction between ``use'' and ``mention'' in (p16-22) is just not present in RDF reification. (I've used "eric" above instead of <rdfprimer> to make the point clearer.) - It is not illegal in predicate logic to have effective(editor,eric) (p24), just illegal in some predicate logics (such as the first-order predicate calculus). - There is no way that RDF quantifies anything, so saying that ``RDF reification allows one to quantify over expressions (statments) and stay inside RDF'' is just plain wrong (p26). The section also has other severe problems: - The section mixes formal notions and intuitive notions to the detriment of both. For example, in (p7) it talks about denotation (a formal notion) and association (an informal notion). A formalist (like myself) cringes when association is used to ``define'' denotation. A non-formalist reader would probably not understand what denote means and become confused as to what the association is supposed to be doing. - How can there be a ``new resource'' (p8)? When does this new resource ``come into existence''? Perhaps one way to see what RDF reification is not is to take an RDF document that uses reification (and doesn't use ID for resources or bagID) and change Statement to STATEMENT in the document. Has anything about the document really changed? If you think that something important has changed, what is it? If you think that nothing has really changed, then what role was reification playing in the document? Peter F. Patel-Schneider Bell Labs Research Lucent Technologies
Received on Wednesday, 24 October 2001 08:38:09 UTC