Comment by skygazer
13 hours ago
I have an absurd and overwrought system involving Gmail, and client-side rspamd and SpamSieve on my Mac. Gmail is (was?) overly aggressive flagging things as spam, so I have the client-side Bayesian filter check Gmail’s spam folder and rescue good email, so long as rspamd also says it’s not phishing. And then add sender to a Gmail whitelisting rule. All rescued email is flagged such that if I later manually move any of it back to junk, it stays there as spam and updates the corpus.
I now never get good email in the spam folder, and never get undetected spam in the inbox, and very occasionally get a spam erroneously rescued, but still visually flagged as iffy-but-maybe-ham.
If Gmail has been lax at filtering spam lately, I haven’t noticed, but perhaps the Bayesian filter has been picking up the slack.
I should consider this - I run my own domains, and for years I just forwarded it to gmail, but I had so many cases when mails were put into spam, even replies to emails I had sent in the middle of a long conversation between myself and 1 other person, that I went to just self-hosted IMAP. Then for years I couldn't reliably send to google or yahoo or MS; I added SPF a while ago which help, but recently buckled down and put in SRS and DMARC and DKIM (and rspamd while I was at it); now I get the mail I want, and can mostly send mail without it being rejected (still have to ask people to check spam, but anyways many people I have to tell them I'm emailing them anyways if its important). However I have a lot of non-spam "promotion" emails that I don't want to see. If I could train gmail to not block legit stuff reliably, that would be worth trying again (I would say except for the privacy implications, but since so much email involves gmail on one side or the other, they probably get most of it anyways).