
The right file format should make a clear distinction between the text, and how the text looks, and should allow me to /easily/ apply any arbitrary presentation to that text.
I think at least some of us who have done some books have reached the conclusion that authors and publishers of paper books do not magically conform their publishing practices neatly into a finite number of definitive categories to which definitive markup could be applied, and even if they did there are too many categories for volunteers to get it all "right," and/or to care that they get it all "right" and/or that any committee of 12 could ever agree upon. We can't even agree on what a "paragraph" is or isn't for god's sake, or what it should look like. DocBook for example has almost 400 tags, as does TEI. Find me one book from a famous author where the first edition is poorly formatted. Yes many first editions have lousy illustrations. But bad formatting? Nope. I could probably be happy with little more than half a dozen tags -- if they were actually well-supported.