- From: Peter F. Patel-Schneider <pfpschneider@gmail.com>
- Date: Tue, 21 May 2013 22:42:23 -0700
- To: RDF WG <public-rdf-wg@w3.org>
Here are my comments on Semantics. Some of this I guess I'll fix up as editor. Some I may need help on. Some Pat and I will iron out, hopefully. The main sticking point I see is the statements about blank nodes. There was some discussion of how normative is supposed to work now. Does anyone have a pointer to the rules? peter Comments on RDF 1.1 Semantics 1/ Need to fix up printing - too big - notes don't show up right probably copy stuff from Concepts 2/ Need to say what is normative? I believe it is 1 (!), 2, 3, 4 (except summary), 5, 6 (?), 7, 8, 9, 10, 11, 12 (except 12.1), 13. 14 is not normative (and should be removed, see below). 2/ Need to copy the "key words" boilerplate from Concepts Section 2. 1, after first paragraph, copy boilerplater from Concepts 3/ Fix terminology from Concepts 3, first paragraph - add IRI - remove RDF source? 4/ Need to say what the syntax means better 3, last paragraph Throughout this document RDF graphs and other pieces of the RDF abstract syntax are written using Turtle syntax [Turtle-TR]. The namespace prefixes rdf:, rdfs:, and xsd: are as in RDF 1.1 Concepts, Section 1.4. When the exact IRI doesn't matter, the prefix ex: will be used. When .... Angle .... 5/ More care about blank nodes 4.1, first paragraph conventional logical notion, with only a local meaning. 5, on union This isn't right. A subgraph of an RDF graph shares blank nodes with the graph it comes from. Treating it as a separate graph is some weird operation that comes from nowhere. To make this point correctly would require saying that there is an RDF graph available from a source. If pieces of it are transmitted across the web using current RDF syntaxes and turned back into RDF graphs then these two graphs do not share blank nodes and .... The section on merging should read The process of combining .... Merging is typically achieved by changing blank node identifiers ... blank node identifier. 6/ Separate normative and non-normative New normative section in 4, 4.3 Semantic Extensions with only last paragraph of 4.2 7/ Remove value judgements Opions vary .... -> 8/ Be more careful about namespaces and vocabularies 10. and several IRIs with the namespace prefix rdf: 10. normative meanings on other IRIs with the namespace prefix rdf: 9/ Fix claim about rdfs:Datatype I don't see anything that prevents other members of rdfs:Datatype. I think that OWL actually uses this. 11. the class rdfs:Datatype contains all the references of recognized datatype IRIs. 10/ Better example for rdfs:Literal (because numbers might not be recognized) 12.1 "24"^^xsd:string .... string "24". 11/ Remove 14 It is not the case that distinct RDF graphs can't share blank nodes. Applications can do whatever they please. Someone could even write a surface syntax for RDF so that applications could send representations of blank nodes back and forth. This would be stupid, but it wouldn't violate anything in RDF. Concepts makes most of the points about the referent of graph names. 12/ Fix B I think that you need more domain elements than required in this section. 13/ rdf:value How can rdf:value be used for a property that has several values? First of all, properties (as properties) don't have values. 14/ References RDF Concepts really has to be normative. RDF PlainLiteral has to be normative. Turtle has to be normative (so that the syntax has a normed meaning). XML 10 doesn't have to be normative, as it is only used in explanation. The normative meaning goes through Concepts (and then XML Schema). RDF Schema doesn't have to be normative, as it only describes what is defined in Semantics. RDF Primer can't be normative as it doesn't have any normative content. SPARQL Query doesn't have to be normative (if it is even referenced). XML Schema doesn't have to be normative as everything goes through Concepts. Wording changes: 2. impose user-define -> be given 3. Remove Issue 1 (except that Concepts may pick up a definition of merge) 3. identifiers may have to be changed in order to 7. with rdf:langString as their type 7. Such literals SHOULD be 9. {}-entailment (to prevent bad line break) 9. similar literal -> another literal 9. similar triple -> another triple 10. Some consequences of the RDF semantic conditions ... 11. , and so are RDF entailed B. truthvalues -> truth values
Received on Wednesday, 22 May 2013 05:42:50 UTC