You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by "Tan, William" <wt...@ezecastle.com> on 2004/09/30 22:54:49 UTC
Whitelist to improve performance?
My configuration is Postfix 2.1.5 and SpamAssassin 3.0.0. We're using spamc as a content_filter in /etc/postfix/master.cf to call spamd.
My understanding is that the manual whitelist function in SA simply starts the message scoring at -100. Is there a way to have spamc/spamd abort scoring a message if the sender is whitelisted? I'd think that this would improve performance on these messages since rbls, dcc, razor, and pyzor would be skipped.
If not (and this is more of a postfix question), is there a way to use whitelists in postfix to bypass SA (or a content_filter in general)?
William W. Tan
Chief Technology Officer
Eze Castle Integration, Inc.
50 Federal St., Suite 400
Boston, MA 02110
(617) 217-3006
wtan@ezecastle.com
Re: Whitelist to improve performance?
Posted by Matt Kettler <mk...@evi-inc.com>.
At 04:54 PM 9/30/2004, Tan, William wrote:
>My understanding is that the manual whitelist function in SA simply starts
>the message scoring at -100. Is there a way to have spamc/spamd abort
>scoring a message if the sender is whitelisted? I'd think that this would
>improve performance on these messages since rbls, dcc, razor, and pyzor
>would be skipped.
Disclaimer: I'm not a dev, but this is based on my understanding of the SA
code. Justin/Theo/Dan/whoever, please feel free to correct and subtle
details I'm wrong on.
No, such things are generally implemented in the tool calling SA.
By the time SA figures out who a message is from/to, it's already parsed
most of the message headers anyway and the RBL queries are already
in-progress. Aborting the scan at this point saves some CPU time, but not
as much as you'd like. It would also likely result in an ugly kludge deep
in the heart of the rules engine, or a slowdown by forcing this check to
run before the DNSBL checks can start (instead of in parallel like they are
now) .
Really this is SO much better implemented by preempting the call at a
higher layer.
Besides, depending on what headers your MTA drops in, SA might not even
accurately know who a message is being delivered to in the first place.
SA's whitelist features are really a bit of a kludge themselves, but they
exist for those who don't have any other option.
>If not (and this is more of a postfix question), is there a way to use
>whitelists in postfix to bypass SA (or a content_filter in general)?
This I can't answer, but I suspect there is a way.