#haraka

/

      • ultimatt1 joined the channel
      • ultimatt2 joined the channel
      • ultimatt1 has quit
      • mocx joined the channel
      • mocx has quit
      • mocx joined the channel
      • doublemalt_ has quit
      • dave2k joined the channel
      • dave2k has quit
      • JinOz is now known as Jinux
      • ultimatt2 has quit
      • Jinux is now known as JinOz
      • kpsullivan joined the channel
      • mocxy joined the channel
      • mocx has quit
      • lnb joined the channel
      • Kaustubh joined the channel
      • Kaustubh
        How to queue message from node in haraka i.e.without swaks
      • doublemalt_ joined the channel
      • _smf_
        Kaustubh: huh? swaks is used for testing only.
      • Kaustubh
        Yes, I am trying to queue message from my app
      • and not getting idea of how to do it
      • _smf_
        Kaustubh: you need an SMTP client; Haraka is the server-end only.
      • You'll need to use the smtp-transport for nodemailer if you want to use it with Haraka.
      • Kaustubh
        Okay
      • _smf_
        You'll also have to set-up Haraka for SMTP AUTH and TLS.
      • (or use a relay plugin).
      • Kaustubh
        correct me if I am wrong,t means I will have smtp-client to which my app-server will communicate and that smtp-client will communicate with haraka
      • correct?
      • _smf_
        Kaustubh: correct.
      • doublemalt_ has quit
      • doublemalt_ joined the channel
      • doublemalt_ has quit
      • kpsullivan has quit
      • Kaustubh has quit
      • lnb joined the channel
      • lnb
        morning
      • cek
        TypeError: Cannot use 'in' operator to search for 'errno' in null
      • if ((data instanceof Error) || ((typeof(data) == 'object') && ('errno' in data))) {
      • How is the error possible at all guys?
      • wtf, null is.... an object!
      • EyePulp joined the channel
      • ultimatt joined the channel
      • ultimatt
      • _smf_
        ultimatt: what are you unsure about?
      • The mod to plugins.js looks fine to me FWIW - it's probably worth expanding that to disconnect too.
      • ultimatt
        I think it's okay, just wanted to check with baudehlo too.
      • _smf_
        Ok - you'll need to do a merge into that branch too as it's going to conflict with my commits earlier.
      • ultimatt
        yeah, no worries
      • I'll do that after updating all the other plugins to use that hook
      • _smf_
        Ok
      • ultimatt
        "it's probably worth expanding that to disconnect" <-- yeah
      • I'll do that too, if for no other reason, documenting what ends up happening
      • _smf_
        I had one minor niggle with your noop commit; we're going to get questions on here like 'why did my hook abort?' as you removed the text explaining why the hook was aborted.
      • ultimatt
        no I didn't
      • it's in the log message right above it.
      • So now they'll get two log messages
      • first the "client has disconnected" message
      • and the next explaining the consequence
      • assuming they're running at logdebug
      • _smf_
        Ok - missed that due to the refactoring....
      • ultimatt
        np, that file was a bit of a morass to untangle
      • _smf_
        BTW - I like the karma changes; but the docs *suck*
      • The comments in karma.ini really should be in the plugin docs.
      • baudehlo
        or both but then there's a risk of getting out of sync
      • _smf_
        e.g. WTF does this actually mean:
      • ultimatt
        I think _smf_ is the only person I've had ask for them to be in the docs (vs ini)
      • _smf_
        [email protected]_good = 2 if in
      • "if in" ???
      • ultimatt
        it means, +2 if 'all_good' is in karma.pass
      • karma.pass is an array
      • _smf_
        Ok
      • ultimatt
        and if one of elements === all_pass...
      • _smf_
        So the rhs of the @ is what is compared.
      • ultimatt
        yes, if there's no rhs of the condition
      • but since the INI has to have unique keys (b/c of our implementation of .ini)
      • putting it there makes sense
      • but you can also use this syntax
      • [email protected]_old_thing          =  2 if in all_good
      • _smf_ looks confused
      • some INI parsers let you have multiple identical keys
      • ours does not
      • _smf_
        I guess once I finally get around to looking at karma it might make sense.
      • It's far more attractive now the other stuff has been removed.
      • mocxy is now known as mocx
      • mocx has quit
      • mocx joined the channel
      • baudehlo
        it feels like SpamAssassin lite.
      • ultimatt
        for stopping spam w/o FPs, its quite a bit better.
      • the most effective part of it really is the ASN history tracking.
      • so many of the ASNs are used solely for spamming, and detecting those (recently made easier with the all_bad (IP) and asn_all_bad keywords) and adding "extra" karma points just for those makes a very big difference
      • it's amusing to compare karma scores in ES alongside SA and rspamd scores
      • baudehlo
        when we did anti-spam at MessageLabs ASN wasn't that easy to get. Is it easier now?
      • ultimatt
        ASN is incredibly easy to get
      • there's 4 DNS lookup services available
      • you can download the data via FTP from routeviews
      • it's packaged up with our GeoIP databases (which are updated automatically)
      • what I need next is to get at the SWIP'ed parties behind the ASNs
      • so I could, for example, notice that 6 of the ASNs that are persistent naughty actors all belong to one "outfit" and then proactively assign karma to their other 60,000 ASNs (yeah, one spam operator has that many ASNs :-0 )
      • Instead, I'm pulling the "naughty ASN" data that deserves "special" treatment manually from ES
      • I figure a good "next step" is to adding "connection.transaction.notes.train_as_spam=true" to some portion of the mail, and have the SA and rspamd plugins automatically use it to train the bayes
      • baudehlo
        *nod*
      • ultimatt
        filtering traffic by ASN in ES makes it really easy to see each of the spammers "campaigns" and get a good idea of who hired the bastards to send that spam run. More importantly, I can see that most of their attempts get karma scores of -20 or lower, but how did those 4 sneak by with just -7? " And why are SA and rspamd so *awful* at detecting them? (less than 10% accuracy)
      • baudehlo
        Now we ahve the web interface too, here's an idea: cache the email in some cache, and when a link is clicked, train on that email.
      • and put that link in the banner.
      • ultimatt
        then I have to quarantine at that crap
      • I figure If I can regularly train SA and rspamd with some portion of "fresh spam," then both SA and rspamd will get much better
      • baudehlo
        well I guess I should try karma, because I still get all that foreign crap.
      • ultimatt
        > when a link is clicked, train on that email
      • that creates some interesting problems
      • some email appliances now visit every link in incoming emails
      • proactively
      • baudehlo
        well this is for my personal needs. :)
      • another thing I could probably do is run a process watching the Trash folder on my server, and train on things going there.
      • as I only delete spam.
      • ultimatt
        I click the "spam" button in my MU
      • A
      • which moves the messages to the Junk folder
      • baudehlo
        I don't have that in Mail.app
      • ultimatt
        so I've been planning to get a dovecot sieve
      • You can.
      • View -> Customize Toolbar
      • and add the Junk icon
      • baudehlo
        No I mean Mail.app on the iPhone.
      • I only read sergeant.org mail there now.
      • ultimatt
        ahh, right
      • baudehlo
        and it only has a Trash option.
      • ultimatt
        on my iPhone, I use the "move to folder" option
      • and drop 'em in Junk
      • baudehlo
        yeah that's an extra action
      • ultimatt
        true, it is
      • once I have karma training SA and rspamd automatically, I might finally give up on the managesieve rules for dovecot
      • that's *supposed* to do the same thing
      • upon user actions
      • but has been a real PITA to make / keep working
      • _smf_
        ultimatt: I've been doing that here with DSPAM; e.g. training automatically on certain stuff where DSPAM doesn't already think it was spam (I do a two-phase train; classify first and if < threshold, inoculate)
      • It seems to work pretty well.
      • I use a training method called 'Train on near error', so I prompt the users to retrain messages that were nearly errors e.g. < threshold and that keeps things in check.
      • ultimatt
        I'd imagine so. It worked fairly well for me with QP and dspam. I don't yet have dspam for Haraka, so I've been eyeing rspamd's bayes as a shortcut
      • _smf_
        Yeah; it's a shame about the rspamd implementation using mmaped files though.
      • notkoos
        _smf_: it should support redis (and possibly sqlite) in 1.0 ;) it was a target for 0.9 but it got pushed out ;)
      • _smf_
        notkoos: excellent - Redis would be really good (far better than SQLite IMO).
      • SpamAssassin added Redis support to their Bayes and it's insanely quick now.
      • teknix__ joined the channel
      • teknix__ has quit
      • ousado joined the channel
      • ousado has quit
      • ousado joined the channel
      • lnb
        why are we still getting funny characters like G4225 SCu ss     1599.00- discountÂ