- From: Sergey Melnik <melnik@DB.Stanford.EDU>
- Date: Fri, 03 Dec 1999 23:07:06 -0800
- To: RDF Interest Group <www-rdf-interest@w3.org>
RDF IG, thank you all for the valuable input on the RDF API. I hope we could produce something usable by the end of the millenium ;) Below you'll find the current sketch of the core RDF API, mostly based on the work and comments of JanneS, RonD, GabeB, JonasL, MichaelE and others. For convenience, I'm using direct Java notation instead of CORBA IDL. Method names are in parentheses: org.w3c.rdf.model ================= interface RDFNode (empty interface) interface Resource extends RDFNode (getURI) interface Literal extends RDFNode (getString) interface Triple extends Resource (predicate, subject, object) interface RDFModel extends Resource (setURI, size, elements, contains, add, remove, find, duplicate, create, createResource, createLiteral, createTriple) interface SetModel extends RDFModel (union, difference, intersection) interface VirtualModel extends RDFModel (getGroundModel) org.w3c.rdf.syntax ================== interface RDFConsumer (startModel, endModel, addTriple) interface RDFParser (parse) interface RDFSerializer (serialize) class RDFModelConsumer implements RDFConsumer (allows parsing into an RDFModel) org.w3c.rdf.syntax.sirpac (not really core, just an implementation of RDF M&S 1.0 parser/serializer) ========================= SiRPAC + Serializer adapted from GINF The package namespace is structured according to the specs produced by the W3C. As you can guess, org.w3c.rdf.schema is also there (but is hidden for now ;). I know it is hard to discuss the details without looking at the code and trying it out. I'm working on packaging the stuff and putting it on the Web for your critique. If you see something that is conceptually wrong with the package structure or interface hierarchy, or have better names for the interfaces, please let me know. ISSUES: interface Triple implements Resource, i.e. must have a URI. It seems logical that the URI of a triple is the URI of the corresponding reified statement. For some plausible reasons, the URI of the reified statement should be unique. [1,2] propose some algorithms for computing this URI. The advantage of the approach presented in [2] par. 4 is that this URI is still human readable. The disadvantage is that it can get very long (especially if the object is a literal; note that I'm not advocating for adapting [2] for treating each literal as a resource). In my view, this is an unresolved RDF model issue and should be addressed asap, since it affects both the API and the syntax (the triples generated by the parser). PLEASE share your thoughts! Sergey [1] http://lists.w3.org/Archives/Public/www-rdf-interest/1999Nov/0092.html [2] http://lists.w3.org/Archives/Public/www-rdf-interest/1999Nov/0115.html
Received on Saturday, 4 December 1999 02:01:42 UTC