List Mail User wrote: > Do you have any idea (or even better, measurements) of what the FP > rate would be if you 4xx'd it instead of 5xx'ing - i.e. how many of those > FPs are corrected before the MTA re-delivery timeout period (yes, I know > that many sites use less than the recommended 5 days).
I have no idea. However, on a related note, I've always wanted to add a feature to milter-greylist to greylist based on RBLs. DULs, XBL and spamcop would be truly great lists to do a greylist based on. It's been suggested and is on the milter-greylist todo, however nobody with time has had enough provocation to do it. (I myself lack spare time to work on it at home.) Using greylisting you'd delay their mail, but they'd be able to deliver even if they still are in the RBL if they retry after the greylist timer expires. This differs slightly from using a list as a 4xx, in that delivery doesn't require delisting from the RBL. I currently do selective greylisting using IP ranges (apnic, lacnic) and RDNS name regexes (dialup-esq RDNS) and nearly all the mail I greylist never retries at all. I can't afford the delays of greylisting everything, but I could afford it with a RBL that's 99.5% or more spam.