[jarrettc]

I agree that HTML needn't be strictly semantic in the way that the OP suggests. But for the foreseeable future, it does need to encode basic structural data for CSS-less presentation. That includes links, headings, inter-page and intra-page navigation, and the like. Who needs that data? Screen readers, text-only browsers, and scrapers, to name a few.

Someday, it may be possible to do away with the concept of a web "page" or "document." Desktop and mobile apps aren't (generally) built in that paradigm. But if we move away from the document concept, we'll have a new set of problems. How do we represent things like blog posts, which really are more like documents than applications? How do we enable a variety of clients with varying capabilities to consume textual content? How do we maintain the useful concept of a URI, if at all? These may not be unsolvable problems, but they are challenging for sure.