The W3C's Semantic Web Activity Statement is a rundown of what they're doing toward the Semantic Web. Contains the intriguing line "investigations into the use of the XML DOM for non-tree-structured data". Oooooh.
One of the problems with building a vision for the future is Who's going to adopt it? And will all those people will Geocities accounts really start marking up their content in a standard dialect, separating form from content? No, of course not, not as it stands. But half of building the future is building a path to get there, and that's something I think the W3C are doing quite well. They've obviously got some kind of internal vision of the Semantic Web which only manifests itself as technological specs or demo implementations -- the danger is that then somebody sees the obvious next step, and builds a proprietary implementation of it. But agreed open standards that build and work with other technologies appear to win out in the end, or at least I trust that they do. Slowly, slowly, and in the end we'll all be putting together perfectly marked-up pages and won't even notice.
Case in point: Let's see what happens to Annotea, the demo W3C annotations system (the documentation has screenshots). Interesting that they're actually working on [what to begin with I thought was] a lightweight system built on rdf and xpath, with extensibility built-in for multiple comment types and full xhtml pages as the annotations. It's all about the path, man. [see also this MeFi thread].