[DiapDealer]

It would be pretty hard to determine if the source material was ePub3 or just ePub2 with elements of ePub3/CSS3. Or indeed whether an ePub was even used at all to create the Kindlebook. It's all getting pretty gray.

Surely, the OPF (and maybe NCX vs NAV document) is where the only real differences lie between outputting ePub3 as opposed to ePub2, right? We do nothing now (to my knowledge) to filter out any xhtml/css that's not ePub2 "compliant", so I wouldn't think it would out of the question to have two different OPF generation routines and just give the user the option to choose 3 vs 2 based on what they're seeing. Perhaps even the option to forego ePub generation altogether (just the OEBPS output)?

EDIT: it occurred to me that we could be trying to push html5 into an xhtml-shaped box at times, though. But that still seems only a matter of leaving off the DTD and xhtml namespace, no? Specific attributes and links are massaged, but the markup itself should come through relatively unscathed--regardless of whether it's xhtml or html5--shouldn't it?