Matthias Leisi wrote:
If the application's behavior is simple and well-defined, this might
be good enough, of course.
FWIW when I read the original email in the thread I got the impression
that the application behavior was pretty simple WRT this table. But of
course I could easily be wrong...
You are not wrong. The use case is in fact the `userpref` table used by
Spamassassin. Left unmaintained, and given a large-enough user base,
this has the tendency to grow considerably over time, so we want to
gradually remove entries not actually used any more. (And we don’t want
to patch Spamassassin core code to do this by itself…)
For this particular use case, would it be easier to periodically compare
the list of usernames with userpref data against active accounts on the
mail system?
-kgd
