Re: current version of Feature Synopsis document

From: herman.ter.horst@philips.com
Subject: Re: current version of Feature Synopsis document
Date: Fri, 7 Jun 2002 14:05:20 +0200

> In order to facilitate further work on the document of which
> Peter provided a new version (6 June 2002) at
> http://www-db.research.bell-labs.com/user/pfps/owl/owl.html,
> I summarize below the points of my previous comments
> (at http://lists.w3.org/Archives/Public/www-webont-wg/2002Jun/0008.html)
> which are not yet addressed or not yet addressed completely.
> 
> --Section 2
> The term individual-valued property is an improvement.
> However, the ambiguous phrase "Individual IDs" remains.
> Replacement of the term individual by resource would
> simplify terminology.

I resist this change.  The term ``resource'' is, I feel, not appropriate at
the level of this document. 

> There is now a reference to model-theoretic semantics.
> This is a meta-remark. I suggest combining all meta remarks
> in Section 1 (and/or an appendix), so that Sections 2, 3, 4 are 100% 
> devoted to what the document title says: giving a feature synopsis of 
> full OWL. 

I feel that this would unduely complicate the document.  Why not leave the
(small) description of the meaning of the features in-line?

> At this point in the text there still is a gap in the 
> description of the language features and their meaning, which could be 
> closed with sentences like:
> -Classes denote subsets of the set of all individuals.
> -Properties denote sets of pairs of individuals and/or data values.

These are in the last paragraph of section 2.

> --Section 3
> Two remarks about frames.
> Because the language is meant for the Web, it is of interest, also 
> on the abstract level, to be able to distribute axioms for a class or 
> property into different parts.  I suggest to add this explicitly.
> For subclass axioms, meaning is given by intersecting the parts.
> For property axioms, this is slightly more subtle.

I take it that you want wording to indicate the effect of multiple axioms
for the same name.  I have added a paragraph on this just before 3.1.

> Second remark: Use the term frame only in historical/meta remarks
> (to be moved out of the feature presentation flow). 
> Each simplification of terminology reduces the threshold 
> between OWL and its potential users.

Why should there not be an allusion to frames?  This way of organizing the
axioms was chosen partly for its similarity to frames, so I see only
utility in mentioning the relationship. 

> --Section 3.4
> The notion of restriction was considered difficult in DAML+OIL.
> I suggest to replace the first, brief, explanatory sentence by the more 
> explicit sentence: "Restrictions enable the definition of classes in 
> terms of constraints on properties."

I view this sentence as more misleading than what is currently there.  I
have, however, modified the current text.

> --Section 4
> I agree with Jeff that the first production seems superfluous.

It is not.  A fact is either an individualFact, a SameIndividual, or a
DifferentIndividual.  However, inside a propertyValue, only an
individualFact is allowed, fixing a bug that you pointed out.

> The problem I noted earlier still seems to exist, with other
> production names. It is not just a matter of choosing the right
> name, as Jeff suggests.

Which problem?  I fixed the problem that existed before.

> The first line of the <propertyValue> production implies that a 
> <individualFact> always stands for an individual.

Yes, an individualFact is an Individual, i.e., an RDF description element
with a slighly different syntax.

> But the production for <individualFact> says that a <individualFact> 
> stands for one or more assertions about an individual.

Yes, an individualFact can include one or more propertyValues, each of
which provides information about the individual.

> What is meant here? 

Very close to what is meant in RDF.

> Is an individualFact implicitly also an individual (reification), 

NO REIFICATION!!!  NO REIFICATION!!!  NO REIFICATION!!!  NO REIFICATION!!!

> or does the individualFact need to be projected to the 
> individual about which it is an assertion?

I don't understand this.

> In the latter case, the first line of the <propertyValue> production 
> should in my view be omitted, and replaced by the addition of such a 
> projection function.

I don't understand how this does anything besides introduce problems.

> It is desirable to add explanation about how facts about anonymous 
> individuals work.

You mean besides referring to RDF?  I'm not particularly willing to include
enough to make a significant difference in comprehensibility.

> Herman ter Horst

peter

Received on Friday, 7 June 2002 08:56:57 UTC