If the original source you use turns out to have errors, as nearly all books do, do you want the errors preserved?
One of the advantages of an XML (or similar) "master version" is that we can have our cake and eat it too. The MASTER can capture the mistake AND the correction; a separate process can render either or both in various formats, e.g. HTML with little "tool tips" and of course plain text (optionally with or without "errors", anachronistic spellings, etc.). Three examples: <alt why="modern" what="today">to-day</alt> <alt why="typo" what="spelling">speeling</alt> <alt why="intentional" what="for">f8r</alt> As an example of the third case, Twain's CT Yankee includes a "newspaper article" full of "typos" that are intended to illustrate an amateur job of typesetting. NOTE that the xml tags and attributes are just made up on the fly. I'm NOT advocating any specific tags. Also, I think it's better to start with something rather than wait for a perfect design. It's generally easy to transform from one xml into another, e.g. to (or from): <archaic modern="today">to-day</archaic> <typo correct="spelling">speeling</typo> <intentional index="for">f8r</intentional> -- Scott Practical Software Innovation (tm), http://ProductArchitect.com/