Apparently, the Internet Archive is no longer generating Djvu files, so the folks on WikiSource don't have a web interface for generating them any more. We should investigate what would be involved in setting up a simple interface on Tool Labs for generating Djvu files from PDFs or a set of JPGs.
An issue was identified when the JP2 files had different filenames than were expected (e.g. containing [^a-zA-Z0-9]) and this is now fixed. It seems that the page names in DjVu XML are somewhat constrained.
I think the patch is ready to merge. Will wait for some more testing though — https://tools.wmflabs.org/ia-upload/test/
I asked on labs-l and they suggest using a database instead, but that a 0600 json file is still okay. The file is created like this:
$oldUmask = umask( 0177 ); touch( $jobFile ); umask( $oldUmask ); chmod( $jobFile, 0600 ); file_put_contents( $jobFile, \GuzzleHttp\json_encode( $jobInfo ) );