Hi Turbo,
On Fri, 22.12.2006 at 20:42:54 +0100, Turbo Fredriksson <turbo@bayour.com>
wrote:
> >>>>> "Toni" == Toni Mueller <support@oeko.net> writes:
> Toni> On servers I tend, I see regularly about
> Toni> 15 CPU seconds burned _per_message_, which I find way to
> Toni> high. OTOH I'm unaware of alternatives that have comparable
> Toni> filtering abilities.
>
> Personally, I'm about to completely throw out SA. It suck so badly, both
> in performance and the amount of spam that it lets through...
umm, I'm almost happy with the filter capabilities of SA. It seems that
it does not let too much spam through where I'm responsible (ie, well
over 90% success). That's with only static rules and DNS tests.
I have the requirement of accepting emails of wildly varying quality.
The filter method employed must be able to tell chinese spam from
chinese good mail, as well as russian spam from russian good mail, and
the same for a good handful of other languages, all on the same server
(corporate user with a few hundred email users). I also have to cater
for individual mail senders who have (very) broken mailers. I heavily
doubt that Bayseian filters can do that after experiencing how easily
SA's Bayes filter was poisened to the point of being completely useless
with a much smaller spectrum of emails.
> I installed bogofilter a couple of weeks ago. The day after, I got FIVE
> spams in my personal folder (compared to about 200 with SA).
As an alternative or addition to SA, I've started to look at DSPAM, but
I'll probably look at bogofilter again.
> And it keeps being low. I have yet to learn it properly about Asian spam,
Do you also get Asian non-spam, like I do? I'm very interested in how
> And bogofilter is a lot faster, and uses much less CPU/memory...
Naturally. I guess that all database-lookup only systems, and then also
written in C instead of Perl, should be quite a bit faster. The first
line of defense still is to prevent illegitimate email reaching the
filter, however.
Best,
--Toni++
|