Quinn Comendant wrote:
I too am confronting the issues with simscan not knowing the actual destination 
user account. I haven't setup SQL bayes (yet -- I hope to soon) but I am using 
per-user SQL user prefs.

I have been living with this limitation because blocking spam at the SMTP level is very important (IMHO) both for simply keeping 90% of incoming mail off the server, and also for informing senders that their messages were not received (instead of dropping them into a void).
The solution I've been considering is to run two instances of spamassassin: one 
through simscan, with a global bayes db blocking messages scoring about XX (I'm 
using 8 actually), and a second instance executed after mail has entered the 
system (at the mail-delivery level) that uses per-user bayes and prefs. Yes, 
scanning some messages twice. The disadvantages of this I can think of are:

        - more server load
        - more complex administration
        - bayes may not be trained as accurately because most mail will be ham 
(?)

The advantages:

        + SMTP-level blocking of most spam
        + per-user bayes and prefs
        + two layers of SA bayes filtering might catch more spam (?)

Maybe the simcan-level SA can run all the non-bayes tests, and use mailfilter 
to rewrite the  X-Spam-* headers as X-Spam-A-*, while the qmail-level SA can 
run nothing BUT the bayes tests and have its headers rewriten as to X-Spam-B-*, 
the add the scores together into X-Spam-Status. This would help server load by 
splitting the work between the two SAs, and you could see headers for each.

I'm sure a very clever (and simple, as opposed to the above) solution to this 
soon. DSPAM support might help!

Specific comments:

So I found a solution that works. But I'm not sure if there are negative implications in doing this (i.e. Performance issues, dropped emails, etc).

The load on SA will be the same but I think the load on all the other qmail 
processes will increase.

2. Created a .qmail file in my account folder (/home/vpopmail/domains/somedomain.com/someuser/) that contains the following:

Be aware this file will be overwritten or deleted by QmailAdmin when the user 
edits any of their forwards/autoresponder settings!

It would be better to place the file at:
/home/vpopmail/domains/somedomain.com/.qmail-someuser

I tried your suggestion above and it didn't work for me. Do I need to modify the statement below to work in this file? I modified my mailfilter script to point to the correct mail folder for delivery (it uses `pwd` to get the current folder, so it won't work right unless you append the username to the end of the path) which didn't work either. Just asking if .qmail-someuser should have a different format than the one below. Any clue?

|/usr/bin/spamc -u [EMAIL PROTECTED] | /var/qmail/bin/preline /usr/bin/maildrop -A 'Content-Filter: maildrop-toaster' /etc/mail/mailfilter

It sure would be easier if you could just have one global .qmail file. I'm not 
sure how to do that, but you could use $(echo $RECIPIENT | sed -r 's/[^-]+-//') 
in place of [EMAIL PROTECTED]

It seems to work great. I just wonder once I put it into production with over 100 domains how well it will work. I know I'll have to write some scripts to update everyone's .qmail file, but that is fine as long as I know it will work under a load.

Let us know how it works!

Quinn

---------------------------------------------------------------------
     QmailToaster hosted by: VR Hosted <http://www.vr.org>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


---------------------------------------------------------------------
    QmailToaster hosted by: VR Hosted <http://www.vr.org>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to