> > 1. It specifies that, when adding up the sizes of files to compute the > > size of a mailbox, only the contents of the "new" and "cur" directories
> > should be counted, and *not* the "tmp" directory.
>
> That is correct. Normally, the user has no access to the tmp dir (which
> under normal circumstances should be empty anyway), so it probably
> should not count the tmp dir.
If you think about delivery and quota long enough, the conclusion is that
it simply does not matter. I think I could show cases where accounting
tmp is both better and worse than not accounting it. The maildir quota
algorithm is broken for the sake of being lock free, but that usually
does not hurt. You can not fix it by accounting or ignoring tmp.
I vote for a similar mechanism to size determination to determine the ago
of files without stat(2). Account tmp and clean up old files found there,
which does much more towards a mostly correct result than ignoring tmp.
