Uploading the original PDFs to a publicly accessable website would most likely be a copyright violation, so we wouldn't want to do that anyway. However the originals need to be available somehow (pehaps in a restricted sense) for people to verify the OCR against when marking up (I'm assuming that this is to be going on to wikisource), as an error in a formula would be very hard to spot for a layman
Another question is what to do about about diagrams (assuming that there are some), I would imagine that if the the RS claims copyright of the scans we can't just extract them and use them. Simple ones I imagine we can (and probably should) convert to SVG, but for more detailed ones, that could be tricky.
James