Also: bare-ass raw HTML5 is actually really, really good. It does a nice job of providing structural definition, and I strongly recommend Mark Pilgrim's Dive Into HTML5 as both a guide and demonstration (in its Web incarnation) of that ability.

It still lacks a few things, which both hNews and ePub could extend, and a few other bits.

1. Metadata. Authorship, publication date, title, and some sort of content valididty really need to be a thing. hNews and ePub both go there.

2. Notes. I don't care how they're presented, but sidenotes, footnotes, endnotes, need to be integral to HTML. I've gussied up some solutions myself, they kinda-sorta work. I'm not entirely thrilled. These should be entirely browser-native and not require JS.

3. Some sort of hierarchical discussion. Comments, etc. Essentially a marriage of the browser and the Mutt or tin/rtin newsreaders. With filtering and other capacities.

4. Reputation management. Entirely local, or distributed. Good and bad authorship, as with elections, should have consequences.

5. Curation. Why this is lacking from current generations of Browser I don't know.

Add to this an in-browser set of styles which can be applied to the document templates mentioned above, and we're getting somewhere.

People will write shitty websites in straight-up 8-bit ASCII.

If there's not some external authority running herd, they'll just do stupid stuff.

All body text in bold. Or <h4>. Every last goddamned paragraph explicitly positioned. Yes. For real, some guy who writes interesting atmospheric physics stuff in ... Colorado or Utah. The colour choices. The animations. The autoplay audio / video. I could go on.

Any tagging is sufficient rope. People are idiots. Either rank/placement tools which downgrade that, or a client which effectively does on-the-fly rewrites of the entire source, are pretty much what you need.

Oh, and don't get me started on the clever types. You've seen fonts constructed entirely out of CSS, no?