piece from Alex
James giving a relational database view of the Web, in which links
in a page can be seen as foreign keys. John calls it an analogy,
but I suspect it's more than that.
By continuing the hyperlink-as-foreign-key view, there's room in the relational model for expressing the relationship between the current page (URI) and the remote one. Or more correctly maybe that there is a relation containing the two values. If that relation (i.e. a two-column table) can itself be named with a URI, then the picture of Web of Data becomes even clearer. Yup, that's essentially the approach taken in RDF.
PS. A pretty important bit I missed on first typing: there's no reason that the relationships described within a document have to involve that document. As often than not we'll want to talk about other resources and their relationships. Â
The relational model of the SemWeb differs somewhat from Codd's/SQL's, for convenience and of necessity. Binary relations are a whole lot easier to manage and interconnect than n-ary ones, adopting a closed world model would lead to immediate breakage in the global environment (similarly "perfect" hypertext systems are a lot more brittle than those which allow for a (404) Not Found).
When talking databases, there's another point which John captures neatly:
SPARQL is key because it allows people to transfer their database skills to the semantic web.
'Course Alex is talking about the current Web of human-readable content, from a database point of view the text of the page could get relegated to something like /* comments */. But microformats and eRDF demonstrate that such content can be treated as first-class attribute values.@en