Books have their own foibles; table of contents, notes, indexes etc. These really need a semantic approach.

Finally, I'm not suggesting anyone handcode it. I use a writing tool on my iPad that uses markdown as its file format. I never see the markdown, I just bold text, add headings etc. That was the idea, to keep it in the background. Hence the reason I'm aiming for simplicity.

Incidentally, most books need only ten tags, five of which are used exactly once on the title page.

My ultimate word processing environment would be an empty web page where you can type Markdown and have the UX adapt to what you are typing, much like entering comments on StackOverflow. When you are done you could apply global styles or even page templates without changing the semantic content. You could even go back to authoring in the utf8-based language if you preferred.

Outline view in Word is useful, section navigation (and section styles) in OpenOffice are useful. Text to table would be great if it actually worked.