Comments on RDF meaning and machinery
Some Comments on Resource Description Framework (RDF): Concepts and Abstract Data Model http://www.w3.org/TR/2002/WD-rdf-concepts-20020829/ The title of Section 2.2.6 is misleading at best, and just plain wrong at worst. RDF cannot ``say anything about anything.'' For example, it cannot say that every person has at most one parent. So what can RDF say? Does RDF allow ``universal expression of ground facts'', as stated in Section 2.2.7? No, RDF cannot say, for example, that Sue or Ellen is John's sister, or that Sue is not John's sister, or that if Sue is John's sister then so is Ellen, all of which are ground facts. Maybe RDF can say atomic ground facts? No, RDF cannot say, for example that there is an purchase property between John, Susan, and John's pet rock. Is there any place in the document that correctly states what RDF can say? Not that I could find, not even the statement that RDF can say ``assertions of specific properties about specific named things'' because that ignores unnamed (or blank) nodes in RDF graphs. The document says that to support use by automated tools ``certain meanings of RDF statements must be defined in a very precise manner''. Does this mean that there are several, possibly different, meanings of RDF statements? The document says that the RDF core language has ``no machinery for formalizing allowable inferences''. What then is RDF closure as defined in the RDF Model Theory document? The documents says that many of the nodes in an RDF graph are blank and some are labelled. Why the differences? Does this mean that there can be no RDF graphs where all nodes are labelled?
See: http://lists.w3.org/Archives/Public/www-rdf-comments/2002JulSep/0174.html
Various objections to text dealing with assertions and meaning in RDF, particularly at the overlap between formal and social meaning.
Most of these comments seem to address GK's text. Need to check some comments in detail, as precise target referent isn't immediately obvious.
See: 003-ModelTheory.html
See: 016-NodeLabels.html
1. The phrase "say anything about anything" has been a useful sound-bite, but I agree is not most appropriate here. I think saying that "Anyone can make simple assertions about anything" is close enough for the purposes of this section.
2. What can RDF express? Yes, I got that wrong, misunderstanding the meaning of "ground fact". I'll try to make 2.2.7 more precise. 3. Yes, "certain meanings of RDF statements" is not the most helpful phrasing. See issue 003-ModelTheory. 4. Concerning "no machinery for formalizing allowable inferences", the intent was to say that RDF has no way to express these. I'll change it to: "with no way to formally express allowable inferences". 5. The matter of node labelling has been separated into a separate issue, 016-NodeLabels.See: http://www.ninebynine.org/wip/RDF-concepts/2002-10-18/rdf-concepts.html
Comments folded in to publicly accessible document.