There's been a flurry of posts about the concept of Scholarly HTML, ie self referential truly portable HTML for scholarly publication, allowing you to save a web page for reuse.
Well, that's a laudable aim, and would certainly aid the migration of content. However I think we're asking the wrong question here.
Thinking about reuse is the key - how are we going to reuse it? As a website, as a preprint lodged in a digital repository, a paper document, or reading it on an e-book reader etc etc?
What we actually need is an archival format that can easily (and I do mean easily, no standing on one leg with your finger in your ear geekery here) and faithfully (consistently may be a better word) produce documents in a set of derived formats such as pdf epub, odt and the rest. This id different from initiatives such as TEI as they're targeted to ensuring the accurate transcription and representation of paper texts in an electronic format. XML versions of docbook might be a better starting point
The format has also got to be publicly known - and natively readable - so it's easy to write new transforms - and deal with embedded non text media - XHTML anyone ?