process-removals is really cumbersome

[Colin Watson]

process-removals is an important part of our archive tools: we use it to keep up with removals of packages from Debian, which would otherwise build up and leave us with arbitrarily large amounts of unmaintained cruft in the archive. It copes with checking a variety of places for packages that depend on the removal candidate, making it reasonably safe with some care. It is, however, desperately cumbersome and annoying to use. It has to be run by hand from time to time and takes enormously long to check all removals since the beginning of time, so I added an option to start at a specific date; but that means there are almost certainly a large number of old removals that we've failed to notice, which are now taking up space in the Ubuntu archive for no good reason. Furthermore, there are often non-trivial blockers that require work to resolve, but only people who run process-removals (to a first approximation, just me) ever see these.

I think that process-removals should be entirely refactored into a web page like the NBS report or https://ftp-master.debian.org/cruft-report-daily.txt, which categorises the possible removals into those which can be done immediately and those which are (probably) blocked, sorted by age within each category. This would make the whole thing substantially more pleasant, and would make the work much easier to distribute.