2007 is the make-or-break year for the Semantic Web. The specs are done. The tools are in place, and there's still not a whiff of a killer app anywhere to be seen. The Achilles heel of the Semantic Web may well be the complete disinterest of most authors in producing anything remotely approximating metadata for their pages. Search engines have learned to ignore any user-created metadata because honest publishers don't bother with it and dishonest spammers abuse it. Screen readers don't even bother with the limited semantics already in HTML, trying instead to figure out what the page looks like.

If publishers can't be relied on to provide metadata, then where can you get it from? How about from the data itself? Gleaning Resource Descriptions from Dialects of Languages (GRDDL) is the first Semantic Web technology that does away with the notion of publishers generating their own metadata. GRDDL consults XSLT stylesheets supplied by third parties to scrape the metadata off Web pages. The output from these stylesheets are Resource Description Framework (RDF) triples that you can process with the underutilized RDF toolset. Different stylesheets can be applied to different sites as needed. Indeed, different consumers can use different stylesheets that provide the information they find of most value.

It's a clever plan, and it seems like it might work (which is more than I can say for most of the Semantic Web). But this is the last chance. If GRDDL can't make the Semantic Web happen, nothing can.