On 08/11/2014 02:02 PM, Alex wrote:
Hi, Hopefully you'll consider this a related question, as I would really appreciate your input. We periodically have users that complain about receiving email they believe to be spam, but it looks to be legitimate.
I'm still pretty much a newbie after only 3 months of getting back into administering a mail server. But I'm finding that it's best to consider anything at all legit to be ham, where "anything at all legit" means that it looks legit enough that the "unsubscribe" link would likely work. Even if it's a sleazy "opt out" sender.
SA is sometimes smarter than I expect. And I've only recently discovered the included DNS Whitelist rules. Personally, in my own account, I sometimes get lazy and try to use SA's Bayesian training via dovecot-antispam as a substitute for doing an unsubscribe. But if the email is legit enough to be unsubscribed from, unsubscribing is the best way to handle the situation. And that's what I'm telling my users. That way, bayes can concentrate on real spam, and dns whitelist rules don't work at odds with bayes.
My post may or may not be only be tangentially related to the topic. But I figured I'd mention my recently formed definition of spam. There's a lot of complexity embedded in the SA standard rule set. I try not to make too many assumptions.
-Steve Bergman
