OK, now that things have settled a bit I'll start to look at the implementation of the various filters.
I'll start with my favorite - ASCII85. I'll have to re-read the notes on that since it's been so long since I've read them, but despite that fact I have some suggestions to make to the decoding part: 1. Wherever problems are encountered in the stream being read, please send a message to stderr including the usual GNU coding standard bits (file name, line number), some descriptive text and ,if available, the file position and stream information (location in the PDF, position of the fault, anything else useful). 2. Check for the "end-of-stream" marker '~>'. If it does not exist or part is missing, this must be reported as with any other error. Such reports can help identify non-compliant PDF writers and also help with debugging pdf writers. Don't worry if in the future this results in a PDF viewer spitting out volumes of messages - that would simply mean there are some pretty crappy writers out there - after all, stderr can always be sent to /dev/null. - Cirilo
