Not an issue - RSpamd config

[MrSleeps]

I swapped out SpamAssassin for RSpamd a while back, would it be worthwhile for the project if I post up my config (well, a basic version of it).

Very few changes to Exims config, our spam catching rate improved massively within a week of swapping away from SpamAssassin.

It's heavily tied into our systems but if you think it's worth while I can do a pull request with a basic config that will get users up and running.

Why did I swap to RSpamd? It's more efficient, uses way less memory, it's non-blocking, uses LUA so can literally plug it into anything (handy way to monitor email throughput)

Here's a comparison (although, it is from rpsamd themselves)

[runout-at]

IMHO yes. Because we have spam score values in the database and it would be interesting how to use these with rspamd.

Is there a way to use both? Eg if the result of rspamd is just at the edge of refusing a mail then ask spamassassin too? Would such a configuration make sense? Or are the results too similar?

[MrSleeps]

In theory you could do that but I'm not sure it'd work (nor can I see the point? Double spam checking?). The way we run it is via calling our internal API to get the user scores map (just a tiny bit more overhead than direct sql), I'm sure you can do it another way but I am also in the process of untangling our web interface (there's a lot of business logic stuff in there that isn't needed) and api to release it (runs on bog standard php).

It scores differently from SpamAssassin (we generally find tag at 6 and reject at 18 works well) but we chuck anywhere from 30k to 50k messages through it every day of the year and our false positive rate is currently at 1.7%. (assuming the customers properly move things out of the spam folder)

It requires a couple of changes to exims config but other than its pretty much the same.

RSpamd also handles our virus scanning, it was genuinely one of the smartest moves we made swapping out SA for RSpamd.

[MrSleeps]

Some configs are here