I'm particularly interested in hearing from Ms. Lofstrom with suggestions about what WEM metadata should be collected, and how it might be structured and retained.
I would also like to suggest that in keeping with the PG charter to preserve "books" not "parts of books" the "WEM" data (not implying literally the WEM data) needs to be part of the distribution, at least that part of the distribution going out of PG to the end customer. In the case of HTML (used only for an example) that would require PG layering even more PG traditions on top of HTML, because HTML doesn't contain this stuff. It could go at the end, and it could go in in a "hidden" manner (except that that which is hidden is often not hidden as we unfortunately keep rediscovering in the case of pagenums) or it could go in the body in sensible places, such as part of the title page, if there are suitable tagging conventions.