From: "Palvelin Postmaster" <cgpsa-discuss@mail.tffenterprises.com>
Received-SPF: pass
 receiver=mail.tffenterprises.com; client-ip=83.150.109.27; envelope-from=postmaster@palvelin.fi
Content-Type: text/plain;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable
Mime-Version: 1.0 (Mac OS X Mail 11.3 \(3445.6.18\))
Subject: Re: [CGPSA] Delivering spam to Junk mailbox of each recipient
Date: Tue, 29 May 2018 12:52:27 +0300
References: <list-19440160@mail.tffenterprises.com>
To: CGPSA Discussion List <cgpsa-discuss@mail.tffenterprises.com>
In-Reply-To: <list-19440160@mail.tffenterprises.com>
Message-Id: <B7EE6400-E41C-4BAB-9D23-834D3F7F7A7C@palvelin.fi>


> On 29 May 2018, at 10:53, Patrick Sneyers =
<cgpsa-discuss@mail.tffenterprises.com> wrote:
>=20
>=20
>=20
>> Op 29 mei 2018, om 09:09 heeft Palvelin Postmaster =
<cgpsa-discuss@mail.tffenterprises.com> het volgende geschreven:
>>=20
>>=20
>>> On 29 May 2018, at 9:37, Patrick Sneyers =
<cgpsa-discuss@mail.tffenterprises.com> wrote:
>>>=20
>>>> Op 29 mei 2018, om 08:17 heeft Palvelin Postmaster =
<cgpsa-discuss@mail.tffenterprises.com> het volgende geschreven:
>>>>=20
>>>> Hi,
>>>>=20
>>>> I have a system-wide rule which delivers all recognized spam =
addressed to all domains and recipients to one centralized spam folder =
(of an account created exclusively for this purpose) for temporary =
storage and teaching SpamAssassin=E2=80=99s bayes filter.
>>>>=20
>>>> I=E2=80=99m considering changing my strategy to having a per-user =
spam folder instead. I=E2=80=99m thinking this could, at the very least, =
empower the users to actually receive these messages and manage possible =
false positives.
>>>>=20
>>>> Does anyone have any experience or insights about the strengths and =
weaknesses of alternate handling strategies?
>>>>=20
>>>> What would be the best way to implement this? Do I just need to =
make a =E2=80=99Store In=E2=80=99 domain-wide rule for each domain =
instead of the single system-wide rule I have now?
>>>>=20
>>>=20
>>> You can use this option in cgpsa.conf (line 552 in my file), and =
keep using the server wide rule.
>>=20
>> Ahh=E2=80=A6I had forgotten about that. Thanks for the reminder! :)
>>=20
>> So, here=E2=80=99s my current server-wide processing rule. Do I just =
need to delete the =E2=80=99Store In=E2=80=99 action and tweak the =
cgpsa.conf dma settings?
>>=20
>> Subject is *SPAM-PALVELIN*
>> Header Field is X-Spam-Flag: YES*=20
>>=20
>> Add Header
>> 	X-Spam-Global: YES
>> Write To Log
>> 	Spam (GLOBAL) Detected
>> Store In
>> 	~spam/Junk-SpamAssassin
>> Discard
>=20
> After setting dma in cgpsa.conf (and restarting extfilter), you can =
disable that rule completely. Just keep the one above, with the =
"ExternalFilter =3D  cgpsa" action.

=E2=80=9D=E2=80=A6the one above=E2=80=9D?


>>> Of course, learning spam/ham becomes a bit more complicated then.
>>=20
>> Do you mean in a technical or practical sense? I currently have a =
script which learns spam daily from all folders which follow specified =
naming conventions (Junk*, Spam*, etc.). I don=E2=80=99t see a need to =
make any changes to it.
>=20
> Then you're all set!
> Would you be willing to share that script? :)

Sure, but this is very crude. :)

#!/bin/bash

maildir=3D"/var/CommuniGate"

# Learn all mbox's in CGP default locations where name indicated a spam =
mailbox

for SpamBox in $maildir/Accounts/*/Junk*.mbox;
        do
                /opt/local/bin/sa-learn --spam --no-sync --mbox =
"$SpamBox";
                echo "Processed spam mailbox $SpamBox";
        done

for SpamBox in $maildir/Accounts/*/Roskaposti*.mbox;
        do
            /opt/local/bin/sa-learn --spam --no-sync --mbox "$SpamBox";
            echo "Processed spam mailbox $SpamBox";
        done

for SpamBox in $maildir/Accounts/*/Spam*.mbox;
        do
            /opt/local/bin/sa-learn --spam --no-sync --mbox "$SpamBox";
            echo "Processed spam mailbox $SpamBox";
        done

for SpamBox in $maildir/Domains/*/*/Junk*.mbox;
        do
            /opt/local/bin/sa-learn --spam --no-sync --mbox "$SpamBox";
            echo "Processed spam mailbox $SpamBox";
        done

for SpamBox in $maildir/Domains/*/*/Roskaposti*.mbox;
        do
            /opt/local/bin/sa-learn --spam --no-sync --mbox "$SpamBox";
            echo "Processed spam mailbox $SpamBox";
        done

for SpamBox in $maildir/Domains/*/*/Spam*.mbox;
        do
            /opt/local/bin/sa-learn --spam --no-sync --mbox "$SpamBox";
            echo "Processed spam mailbox $SpamBox";
        done

# Centralized spam repository

for HamBox in $maildir/Accounts/spam.macnt//Ham-SpamAssassin.mbox;
        do
            /opt/local/bin/sa-learn --ham --no-sync --mbox "$HamBox";
            echo "Processed ham mailbox $HamBox";
        done

# Learn all mbox's in CGP default locations whose name matches 'Ham' as =
ham

for HamBox in $maildir/Accounts/*/Ham.mbox;
        do
            /opt/local/bin/sa-learn --ham --no-sync --mbox "$HamBox";
            echo "Processed ham mailbox $HamBox";
        done

for HamBox in $maildir/Domains/*/*/Ham.mbox;
        do
            /opt/local/bin/sa-learn --ham --no-sync --mbox "$HamBox";
            echo "Processed ham mailbox $HamBox";
        done


# Its faster to rebuild after all the spam/ham has been fed instead of =
doing
# it every time we feed a spam to spamassassin's sa-learn.

/opt/local/bin/sa-learn --sync


--
Palvelin.fi Hostmaster
postmaster@palvelin.fi