Closed djohle closed 6 years ago
Why not just whitelist the recipients in the SpamAssassin config with all_spam_to?
A fair question.
It's actually been many years since I originally made this change (just now got around to formalizing & submitting it) so I can't remember all the details. However, I think we had tried the all_spam_to approach first, but ran into some edge cases where avoiding scanning altogether was necessary. There is also a potential performance benefit by not invoking spamassassin at all, but perhaps all_spam_to entries get somewhat optimized by the spamassassin engine anyway?
Another example that comes to mind is addresses set up for users to forward definite spam/ham to that are used to help bayes training. We don't want SA interfering with those in any way.
Speaking of bayes, I think autolearning may happen still (as ham) for very negatively scored messages that were in fact spam. I know, that is what bayes_ignore_to is for, but one may not always remember to add an address under both lists.
And there's always that one person that wants to "opt out" of filtering because they are so paranoid that they'll miss something important that they'd rather deal with all the extra junk (seriously, this has come up a few times).
Or maybe we just don't like someone and want to open the flood gates on them for a day >:-) Just kidding, of course!
I have added a feature that allows specifying a list of addresses which will be exempt from scanning. This is useful for certain recipients that need to receive all messages regardless of content. Typically this is special purpose accounts such as abuse@example.com or other internal-use addresses.
P.S. Sorry about all the changes due to whitespace-only edits. I tried 3 different editors and no matter what I did, I still ended up with these extra "differences" in the end. As such I was able to least made separate commits for whitespace vs. actual code changes.