is anyone on the list familiar with Fedora3 and developing content models for TEI documents?

Most examples I have seen (all Fedora2) store the TEI document as a file within a book content model and then the page images in a page or image content model. I'm wondering if anyone has stored TEI encoding at the page level ... with the idea that a script could then combine the encoded pages into a valid TEI document.

This may be not be practical at all.