I run Dovecot and Exim on Debian, which I’ve run (on-and-off) for over twenty years, including at a mid-level agency (hundreds of domains).

One Canada VPS for me (Colocataires)[1] and one UK VPS for my brother (Lagrange Cloud).

I had a complicated qpsmtpd set up for a while but I found that a DNSBL and SPF verifying catches almost all of the email at the SMTP-session stage. I don’t currently run anything after this, although I’m seeing so many shout-outs for rspamd that I will check it out.

I recently enabled full-text search in Dovecot and that’s been a significant improvement.

I used Stalwart previously but the upgrade process was spiky now and then and I appreciate that my Maildir on a disk has been a stable format for decades now. I like their business model and wish them well, though.

I also use Migadu for some domains and they’re great (although not self-hosted, obviously) but using aerc as an MUA means that I want to be physically close to my email provider as it makes latency much more obvious.

[1]: I am a cofounder of Colocataires, trying to not self-promote here, but it’s also a fact that’s where my email is.

score 1 · canonical

I find them highly effective, to the extent that I don’t run a separate spam classifier and my domain is like 20 years old. My unwanted email comes in three forms:

• Hacked residential IPs and proxies; generally either on an RBL but definitely failing to come back after any greylisting and often speaking broken SMTP or sending too many recipients on one email.
• Hacked or cheaply-bought mail servers; again, bad reputation, and if they’re forging the From: they fail on SPF etc. pretty quickly
• People using legitimate mail services (like Buttondown or Mailchimp) to add me to a spam “mailing list” and ride on the reputation of these services. These have been the hardest to stop and will probably drive me to set something up, but I don’t want to accidentally end up classifying like … newsletter services as spam.

score 1 · canonical