Various concerns, especially with social meaning
Resource Description Framework (RDF): Concepts and Abstract Syntax Editors' Working Draft 12 December 2002 The notion of what is and what is not normative in this document is very poorly specified. I have been told several times that the definition of an RDF graph in the document is normative, but yet this section is not marked as being normative. I suggest that each section of the document be clearly marked as normative or non-normative. The document contains quite a number of errors and misleading pieces that need to be fixed. It also contradicts itself in a number of places. Abstract I seem to remember learning that ``This document'', etc., are not to be used in an abstract. I only mention this here because the abstract is otherwise so good, being concise and to the point. Its only other flaw is the superfluous paragraph division. After the excellent abstract, however, this document commits the cardinal sin of repeating the abstract at the beginning of the document itself. Section 1: After the excellent abstract, however, this document commits the cardinal sin of repeating the abstract at the beginning of the document itself. If you say the same thing twice, you should at least have the decency to say it in different words. Just what is the RDF core? Is there something else to RDF besides the core? If so, what, and which W3C documents address it? If I was going by the document list, I would be forced to conclude that the semantics for RDFS were in the core, but RDFS itself was not - a very strange state of affairs. RDF datatyping would fall within the core, which is again strange because it depends on RDFS. I would be careful to spell out RDF wherever it is referred to, and avoid short forms such as ``the formalism''. The notion of layering here is also very strange. There is a distinct difference between the relationship between RDF and DC and RDF and OWL. Section 2: This section starts out with a very clear definition of RDF: ``RDF has an abstract syntax that reflects a simple graph-based data model, and formal semantics with a rigorously defined notion of entailment providing a basis for well founded deductions in RDF data.'' If only the RDF documents actually followed this definition of RDF, instead of swerving out to include ``intended meaning'', ``social meaning'', and the like. The list of uses in the motivations section has just about the best example of non-parallel construction that I have ever seen. The list of design goals is, however, a very close second. It wouldn't take much effort to make these lists infinitely more readable. In the list of goals, there is, again, the statement that RDF is supposed to be able to support assertions about anything. This goal runs counter to several efforts to restrict the ability to state assertions about particular resources. The list of goals also has the very strange goal that RDF is supposed to be a basis for legally binding agreements. What part of RDF is going to be able to support this goal? Why not give up on the term ``data model''? It just causes confusion when compared to model theory. Why are literals in RDF if ``URI references are used for naming all kinds of things in RDF''? Inconsistency is a technical term. It would be best not to use it in informal settings, instead talking about false or incorrect statements. As RDF has only a very limited notion of inconsistency, and then only with respect to datatypes, it is misleading to say that applications build upon RDF have to be able to deal with conflicting information. Why not give a quick glimpse of what a simple fact is here instead of using a bare forward reference? Section 3: Are the concepts in Section 3 supposed to be exhaustive in some sense? It almost seems as if they should be, yet there is no indication of whether this is the case. The RDF Primer says that RDF graphs are labeled directed graphs. This document says that the ``underlying structure of any expression in RDF can be viewed as a directed labeled graph''. Neither is correct because RDF graphs are not really labeled directed graphs as they are usually defined, but one would think that at least the same term could be used. Just what is an RDF graph? As this is supposed to be the normative definition of an RDF graph, one would think that an absolutely correct definition would be given here. However, instead one finds that an RDF graph is a set of triples, followed by a diagram that doesn't look like a triple, followed by a quasi-definition of something called a ``property arc'', followed by the claim that an RDF graph contains statements. Something infinitely better is needed here. The next section talks about the nodes and arcs of an RDF graph. However, an RDF graph contains either triples or statements, neither of which are known to have nodes or arcs. This section of the document also talks about the abstract syntax for RDF, which deserves considerable explanation here, but is not even linked to the appropriate section of the document. The datatype section is explicitly tagged as being normative yet it does not specify which XML Schema datatypes are unsuitable for use with RDF, merely mentioning one that is unsuitable. Similarly, the section mentions that XML Schema Datatypes provides an extensibility framework without specifying how such datatypes can be referenced and there are known problems here. (This is particularly frustrating from my view as I just went through this exercise with respect to OWL.) The section on literals makes the claim that anything that can be represented by a literal can be represented by a URI reference. This claim deserves some support, and support based on the RDF model theory. The RDF model theory does not make recommendations. Instead, it states what RDF means in a formal sense. This is not a recommendation, even if one considers W3C standards as recommendations. A two-place predicate is not a simple fact. In fact, it is not a fact at all! A ground atomic term consisting of a two-place predicate and two argument terms might be considered as a simple fact as can be represented in RDF, but not the predicate itself. RDF really only has the power to represent the binary existential-conjunctive fragment of predicate calculus. The existential-conjunctive fragment of first-order logic includes non-binary predicates, which can only be encoded in RDF, and functions, which cannot even be encoded in RDF. The document earlier makes the statement that RDF is supposed to be able to support assertions about anything. It now goes on to contradict that statement, saying that ``[c]ertain URIs are reserved for use by RDF, and may not be used for any purpose not sanctioned the RDF specifications''. These cannot both be true. This section also implicitly makes the claim that RDF Schema is part of the RDF core, contradicting an earlier claim to the contrary. Section 4: What is the ``social meaning'' of RDF? Does it have any relationship to how an RDF application should act? If so, what is this relationship and how can it be conveyed to an application? If not, what business does this have in a document about RDF? How does an RDF expression get to be asserted? What syntax can I use to assert RDF expressions, or to prevent their assertion? Can I use this notion in OWL? If not, then what good is it? Without any method given for asserting an RDF expression or graph, what good is a paragraph that starts ``When an RDF graph is asserted in the Web''? Maybe this section on social meaning has a place in some commentary on the use of RDF, but it certainly doesn't have any place here. The idea that RDF graphs contain ``defining information'' that is opaque to logical reasoners is ludicrous. An RDF graph is simply a set of RDF triples. It is certainly possible that there can be communities that have intended meanings for these RDF graphs, but these intended meanings are external to the RDF graph, and, indeed, external to RDF as a whole, and thus have no place in a normative part of a document about RDF. What social conventions surround the use of RDF? Even if there were some, why should they make their way into a normative section of an RDF document? The idea that some owner of a URI reference can control the use of that URI reference goes counter to the bedrock goal that RDF allows one to say anything about anything. The RDF model theory contains no hint that any of these sorts of restrictions are possible. This section further reinforces this point when it says that any document found by dereferencing a URI reference has no impact on RDF. The example brings forward these problems. The document at http://skunk.example.org/ does not entail anything derogatory about C:JohnSmith, which is reinforced in the section just above. This being the case, there is no reason for any notion related to RDF to bring this forward. If, however, the opposite was the case then there would be no way for any organization to deploy any RDF-based application. Such applications would not be able to understand the social meaning of the RDF they created or manipulated, and thus could easily create documents holding the organization liable for just about any imaginable consequence. In this case I would have no choice but to tell Lucent Technologies not to deploy any RDF applications. Section 6: Just what is an RDF graph? Earlier it was the Graph Data Model and now it is the RDF abstract syntax. I really do expect a much higher level of internal consistency here. The definition of RDF triple is not much better here than before. Two graphs are RDF equal iff they are the same. It would be much better to call the relationship defined here equivalence, or, even better, isomorphism. The treatment of typed literals here does not match that given in the RDF model theory. In particular, there is no provision for the special treatment of rfd:XMLLiteral there.
See: http://lists.w3.org/Archives/Public/www-rdf-comments/2002OctDec/0297.html
Various points raised. Most fundamentally, concerns with the handling of social (intended) meaning and use of RDF.
Some confusion arounf normative vs non-normative content. Not clear about XML schema datatypes to be supported. Some problems of consistency with other documents. A number of points of terminology, and editorial matters.See: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0275.html
See: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0277.html
See: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0278.html
See: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0285.html
See: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jan/0286.html
See also this thread from the WebOnt discussion list.
See: http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0192.html
Further comments to RDF-comments.