HTML: More structural than semantic
Semantic markup is getting a lot of blog coverage at the moment, following a starter post by Jason Kottke. There’s some great content flowing around (Dave Shea, Doug Bowman and Paul Scrivens in particular devote whole essays to the topic) but the central point is the same: just because a page validates doesn’t mean it’s good HTML; semantic markup is equally if not more important for building good pages.
I’m not about to disagree, and there’s not much I can add to what’s already been said. Instead, I’m going to be picky about terminology. In my opinion, HTML is far more a structural language than a semantic one. The word semantic suggests meaning. Sure, marking something up as a paragraph or header is more meaningful than leaving it in a semantically uninteresting div or span, but to be truely semantic the markup would need to tell us what the element actually is —a headline, or the author of an article, or a list of navigation options for the page. That kind of information is the realm of XML. Semantic HTML elements add a small amount of meaning, but a great amount of structure—a fact that becomes particularly important when you start applying CSS. Try to derive real meaning from the simple elements provided by HTML and you’ll quickly run in to a brick wall.