Is it naive of me to think that subsequent "book releases" of the software would be best encoded as patchsets against the original release? It seems wasteful to re-print and then re-OCR the entirety of the codebase.

While I was typing that, I thought of a second point: would it still be a munition if it were only diffs? E.g. Could one put the patchsets on a public server without legal woes, because it's only the full source that's problematic?