I would tell beginners: Don't define the way different parts of your content are getting displayed, but define what these parts represent. An example structure: Header1, Header2, Content1, Content2, Image1, Image2, Table1, Table2 etc. The author could centrally define and alter the style of each class. As long as the structure is well defined it can easily be exported into xml and be mapped to pdf, docbook and maybe even word etc.
Hey -- tht's a holy grail of SGML, is it not? And that seems to be *exactly* the thing that tends to repell Joe The User. :) Joe really wants to be able to mess with font settings dilog -- sure Comic Sans looks much better tonight than plain Times, and Garamond Italic looks ah-so-much-sweeter than any breed of Helvetica to his eyes.
So, how to convinve him to only care about *structure*? If I knew -- I'd get most people to use something like XXE for most of their document editing.
This really does look/sound like an interesting thing to do, and getting some sort of neutral XML document storage, that could be displayed for editing in some way easy for a user... HTMLArea or RTE -- both look quite nice in that respect. Of course they both are out of the question for the 4.x browsers... But so is much of CSS2 (and chunks of CSS1).
