Sieve with Spamassassin && virus checking mechanisms?

[Rob Siemborski]

On Fri, 4 Jul 2003, foobar wrote:

> Is this (see subject) possible in anyway per account basis in sieve? ( I
> don't remember if there was any proper plugin-support in sieve (haven't
> investigated the code too much ).

No, the sieve code is much less extensible than we'd like (it's hard to
add a new core extention, let alone a plugin).  This is mostly because
sieve extentions can touch basically any part of the definition freely.

> Thing what I done was when I didn't invent anything else and was lazy to
> read the code:
>
> MTA lmtp-> sa_lmtpd (pre-loaded spamassassin-instance which can be re-used
> (better than that idiotic pipe-wrapper) lmtp-> MDA.
>
> by tweakin transports, there is sa.foo.bar which destination is
> lmtp:where_sa_lmtpd_is and every user which wants sa-processed mail
> inserts forward user at sa.foo.bar and it goes thru sa_lmtpd to cyrus's
> lmtpd. That sa_lmtpd is my written perl-daemon which re-uses
> checker-instances.
>
> So,
> Is there any better way to do this?

We use MIMEDefang, which calls SPAMAssassin via the Sendmail milter
interface.  This then inserts a header which sieve can detect and file
into a spam mailbox.  Users can then turn it on and off based on a sieve
script.

Offhand there's no reason this can't do more detailed per-user
configuration except we haven't set up an interface to administer that for
ourselves.

-Rob

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Rob Siemborski * Andrew Systems Group * Cyert Hall 207 * 412-268-7456
Research Systems Programmer * /usr/contributed Gatekeeper