First there were the semantics, and then there are ways of rendering semantic information using things like HTML, or so established wisdom says. That’s why we have XSLT (to transform XML into HTML, for example, not the reverse), code and documentation generators (from UML, or from Java source code), and things like address book applications that can take machine-readable VCARD files and present them in a way that makes sense for humans.

Phil Windley now points to a range of new approaches that start with HTML, and add the semantics. Well, that’s charitable, but they do add enough tags and links (mostly using the HTML tags class, rel, and href), so that a piece of software could reconstruct the semantics it started out with from the HTML (except, of course, that this approach doesn’t start out with the semantics).

For somebody who’s been doing semantics for a dozen years now, this is fascinating. I have no idea whether this HTML-as-semantics will really work in practice, but this might allow us all to express semantics and presentation in the same document, using the same statements!

Technorati tags uses a very similar mechanism to derive a common vocabulary at least…