Jim King is currently delivering a keynote here at Hypertext '04, discussing pdf as a format case study.
He's jumped right into the inside of the pdf file format. This is something we don't do very often, perhaps because it's so close to implementation and we like to take implementation as given. It's hard, in the context of a scientific conference, to talk convincingly about implementation: it's hard to know what it right, and what is simply what we did.
I find this especially interesting right now, because I'm very interested in the problem of planning architectures for Tinderbox and for Storyspace III that will facilitate research collaboration. One approach, clearly, is simply to make the file format sufficiently clear, flexible, and easy to read and to transform. Hmmmm....