Raw

Peter Herndon has posted a great comment in response to the recent TimBL interview (and a bunch of other stuff I linked to), quoted in full below. I thought I'd be able to cover most points with links to the ESW Wiki FAQs, but that'd need a lot of dredging and I need to get some code done right now, so if anyone can help with these points in the meantime that'd be grand. (I've numbered the paragraphs to save quoting).

[1] The basic value premise, as far as I see it, is something like this: Put an RDF layer on top of your existing data sources, and computers that would otherwise have to understand your data’s format specifics can now usefully interact with your data generically. Is that accurate? Or, perhaps I should ask, in what ways is that statement inaccurate? What am I missing?

[2] Something that is interesting to me personally is Bill de hOra’s statement that he’s tired of writing ER specifications, tired of building database schemas, and he says he no longer needs to thanks to RDF. In short, huh? How is he getting away with that?

[3] Here’s my understanding. Now, in a relational database, (speaking very informally here) a given table defines the properties of each row. So, each row fulfills the relationships defined by the table definition — row A has property X, property Y and property Z, where X, Y, and Z are declared in the table definition to be of specified data types. Relational theory 101, bear with me. An RDF triple says much the same thing, at a more decomposed level, if I understand correctly. A triple basically says, something A has a relationship B with something C, where A, B, and C are URIs. Is Bill going to be storing these RDF triples somewhere (a “triple store�, from what I’ve read), and assembling useful data as needed by running a query on the triple store? In what form will that useful data be, a set of triple statements?

[4] Where do ontologies fit in? I don’t think I understand their place in the stack. That is, I understand XML, and I understand schemas, XSD or RelaxNG. Schemas define the syntax of an XML document. OWL/ontologies (are they interchangeable?), do they work like schemas? Can you “validate� an RDF/XML document against an OWL document? And what does that mean? What does it get you? What are the practical implications? On the human level, an OWL ontology assigns semantic meaning to an RDF statement, right? So, a computer can reference an OWL ontology and make meaningful use of it by doing stuff like equivalence operations: “all subjects that have relationships X with objects Y as defined in OWL Z have the same property� — is that a start on it?

[5] I speak Java and Python, with a strong preference for the latter. What are the tools for working with the Semantic Web stack? That is, what are your recommendations? From briefly perusing pragmatron.org, I’m seeing Redland as your triple store, and SPARQL as your query language. Uche, on the other hand, offers 4Suite et al. and uses VERSA for querying. If I had existing relational databases, what would I use to RDF-enable them?

[6] In the interview, TimBL states that certain layers of the SemWeb stack don’t really exist yet, starting with the rules/logic layer, and that neither specifications nor implementations exist. Is that accurate? Does it matter? As Philip J Eby said, XML is a step forward from Java, but a step back from Python for efficiency of expression. My first instinct is to think that the rules/logic layer is, at least for the moment, best realized through code. I can see where a lot of value could be had from having generic, public-accessible stores of XML-specified rules/logic, but I feel a strong sense of chicken-and-egg syndrome. Or at least, practical infrastructure should be built first. Am I on target?

[7] Speaking of specific RDF vocabularies (is that the correct term?), DOAP has the potential to add (monetary) value to a programming project, by making certain things easier — categorization, relationships with other projects (including dependencies), that sort of thing. RDF-ized calendaring and email data has obvious benefits. FOAF, perhaps, could find a place in an identity infrastructure something-or-other, but for the most part (to my eye) has most of its value in the individual social web context. That is, it enables/builds social web contexts between individuals. I haven’t seen FOAF used in a commercial context yet, where it adds (monetary) value in its own right. In short, it has a large “cool� factor, but little commercial application, as yet, though the potential exists. Are there other RDF vocabularies I should be tracking? I’ve been lurking on the Atom mailing lists, so I’m aware of the unsuccessful (accurate characterization?) attempt to make the format be, um, easily RDFable. Is that a reasonable verb?

[8] And, on what mailing lists are Semantic Web technologies discussed and announced?