This I think is the most important question in search of a better web architecture. You could say it's the HTML vs LaTeX question. I don't mean the "hard to dev

This I think is the most important question in search of a better web architecture. You could say it's the HTML vs LaTeX question. I don't mean the "hard to develop with macros and mountains of documentation and ungraspable errors and imcompatible packages" aspect to LaTeX. But rather the question of introspection. (Think also of bibtex and how to compile a document relying on it) IMHO both approaches are crap and barely work. I don't actually think HTML supports web searches or semantics very well, and obviously it is painful to develop interactive applications in, as well. Same goes for LaTeX, even though it's coming from the other side. I think we should have containers instead, with in-band data (somewhere linked in a for-the-web assembly language) and out-of-band metadata (in various formats, to support search engines or other enterprises).. Granted, site providers could play tricks by serving disagreeing data over these channels, but in a way that's possible even now with HTML.