#haraka

i just submitted a couple of PRs. i am here in case you want me to fix up something

morning

afternoon

hey. So I think we're ready to go with 2.8

oooh

I've been abusing the beta for awhile, so far so good, running on node 6

_smf_: that bug in mime parsing must have been there forever.

I can't believe we never saw it before.

baudehlo: is this the issue darkpixel was seeing?

yeah

" + line" <- that was it?

nice find.

yep.

it was easy once I had an email to work with.

well, relatively.

the mail parsing code is a bit complex.

I had to add a bunch of console.logs to figure out where the fuck the lines went missing.

console.logs - the best debugger ever.

it's the only debugger I ever use.

baudehlo: yeah - incredible really given the number of places I've got it installed.

I thought *maybe* it would only happen with filters (banners) in place, but it looks like it happens whenever mail is parsed.

I suspect what happens is mail clients cope with it by being loose in their interpretation of things.

I wonder if that was the cause of some of the DKIM failures that we've randomly seen too.

However I suspect the replacement of line endings around that area too might not help that either.

I'm working on greylisting today - How does that work with multiple MTAs/MX records? Do I need a shared redis greylist DB?

the resends could easily come back to a new server, I assume, so they need to all reference the same greylist records.

I think

Yes you need redis for that.

I'm considering a read-slave to reduce latency from a master redis store. Writes go to the remote shared master, reads come from the local slave instance.

probably unnecessary. I'll try without for a while.

yeah redis is pretty damn fast.

the network latency/traffic is what I'm worried about. Redis itself though is wonderful.

I throw instances of that around like candy

You'd have to do some serious volume for that to be a problem.

You should run greylisting last after all the other network checks (e.g. DNSBL, DNSWL, URIBL etc.)

However - it would be a problem if your Redis instance were to go down. Hopefully the plugin handles that gracefully.

_smf_: I see the hooks on the greylist.js plugin, and what you say makes sense.

I'll need to run a modified version of greylist.js anyhow, so I'll build in some margin of safety for redis disconnects. It wouldn't be too hard to wire up a read slave while I'm in there... No, must resist the added complexity.

_smf_: you've got some time in with rspamd as I recall - have you been happy with it?

It's OK - requires a lot of work though.

It's quite the moving target at the moment - lots of stuff being added to it. Including greylisting.

I've written a few plugins for it and a bunch of rules. It's very fast compared to SA, but there's some rough edges.