I needed to use an article from the Dictionary of National Biography for a Wikipedia reference, and discovered that the Wikisource DNB was not yet ready for prime time. I hacked together a DNB project as a place to park my article. This evolved into Wikisource:WikiProject DNB. The project made tremendous progress after I dropped off the web for awhile.
After a long break, I think I'll start again. I will start with Edward Osborne, and I will use the transclusion method. This page is the place to start.
After another long break, I returned to work on the DNB. I note that further tremendous progress is being made, so I picked a small project and I am working on it. My project is to fill in missing articles. I'm working front-to-back. This is Wikisource-centric. My earlier work was Wikipedia-centric. Volumes 1-3 and some of V4 were complete. The early parts of volume 4 had quite a few bad scans, and these had (apparently) caused other editors to shy away from the volume. When I find a bad scan, I proof against an alternate source, which I document on the talk page in page space. I do not know how to fix this in the .djvu file in commons. When I find a more-or-less readable scan with a very bad OCR (gibberish or interleaved) I use an alternate OCR but I proofread against our scan. If I ever find an indecipherable page with no alternate, I will go to the Library of Congress and use the paper copy. My progress is slow and erratic. When I make the initial bulk correction in page space, I do not mark the page as "proofread" because I usually work too fast. However, after some time (at least a month) I go back and do a more careful actual proofread and then advance the status to "proofread." I find that a one-month hiatus is needed to let me forget the page sufficiently to do a real second proofread. My hope here is that the status of "proofread" will alert any human or robot editor that work has been done on the page and that a mass change to an alternate text should not be done.