History

ManualsLib has been a great place for manuals that are no longer available from the authoring body. However, some of their manuals are watermarked, presumably to advertise themselves, and to make stealing a bit more difficult. These watermarks are undesirable especially if the original document left the Internet unarchived, and--like other similar watermarks--they make any document look less authentic and professional. I think ManualsLib has no rights to claim any ownership over those manuals, if those watermarks imply so.

The first step is to manually-edit each watermarked document with a text editor. Starting without knowledge on the internal structure of PDF, I searched for the URL attached to the watermarks, which became my starting point for trial-and-errors. The goal was to subtractively-remove its watermarks while retaining its readability with SumatraPDF, which uses MuPDF. My success over time made this mundane, so I decided to write a program to automate the task.

The second step is to translate my watermark detection methods into Java code, mostly by pattern matching. Following some standard practices, I wanted the output to be readable with readers other than those using MuPDF. The first working output failed on Evince (Proppler), and Firefox (PDF.js). Some digging led me to Apache's PDFBox, which was brought in for some debugging clues; and some understanding of the internal structure of PDF.

The third step is to refactor the code to output mostly-compliant PDFs. This involves calculating the cross reference offset (startxref), and removing references to then-removed watermarks. This would be much easier with PDFBox, but redesigning the program was too much, and the watermarking of documents in ManualsLib seemed to have shrunk if not ceased, leaving no urge to do so.

The fourth and final step is to brush it up for public release, and--of course--to compose its website. The original release date of 08/01/2021 was pushed back to work on the then new [EDJC API], later used in this project.