Programmer. I think it's great that we're extending the semantics of what we have now, but we're going in the wrong direction. I agree really strongly with John Allsopp that we shouldn't be extending the number of elements available - we should be creating a framework for extensible semantics (using attributes or the like):
I have to disagree here. It is the common document elements that represent the greatest improvement, in that they will provide accessibility cues that classed divs never can. It's all well and good for us to gripe (or crow) about what can and cannot be done with the element collection already in hand (in HTML 4.01 and in XHTML), but when it comes right down to it, we have no way of semantically differentiating the various parts of a document. Attributes within HTML or the implied extensibility of XHTML are not really solutions here, since even with namespaces and custom DTDs there is no way to indicate to a user agent the ACTUAL significance of any given class of span or div within the document context. A convention is NOT a standard.
HTML 5 (and its X counterpart) should mark an enormous shift in the usability and discoverability of web pages for users of assistive technologies, for indexers and cataloggers, and so forth. <canvas> doesn't turn me on nearly as much as <header>, <section>, <aside> and <footer> do.
http://www.alistapart.com/articles/semanticsinhtml5
Though I agree that canvas, audio and video tags, and web workers are great additions.