Mailing List Archive

On Sunday 14 March 2004 10:00 pm, Adam Bown wrote:
> Can anyone help with some basic stuff...
>
> I have SA set up, spamd is running and exim (with exiscan) runs all
> messages through it. It's working fine and is filtering messages.
> However, it seems very "slack". It is filtering REALLY spammy messages ok
> but 90% of stuff get's through. How can I tweak the settings? Is spamd
> configurable?

Enabling network tests can also help, but it slows things down.

--
Give a man a match, and he'll be warm for a minute, but set him on
fire, and he'll be warm for the rest of his life.

Advanced SPAM filtering software: http://spamassassin.org

On Sunday 14 March 2004 11:02 pm, Adam Bown wrote:
> Thanks for the reply, it sounds like that is my problem. I've found
> sa-learn and that looks easy enough to use (although I don't have much spam
> to train it with yet so that will have to wait). In the meantime, how do I
> check if it's doing the bayes filtering? ...and if not turn it on?

Bayes, and bayes autolearning, are on by default. If bayes is on, you'll see
a BAYES_XX rule in the rules summary for spam, or in the X-Spam-status header
for non-spam. You can also pipe a message to the command line
"spammassasin", and give it the command line option "-D" to get extra info on
what it's doing with bayes (and lots of other stuff).

--
Give a man a match, and he'll be warm for a minute, but set him on
fire, and he'll be warm for the rest of his life.

Advanced SPAM filtering software: http://spamassassin.org

On Sunday 14 March 2004 11:06 pm, Adam Bown wrote:

> Is spamd generally quite slow?

I can't really give an answer from personal expereince, because I use SA just
for a single email account. I should note, though, that turning on network
tests won't cause much more CPU load; it just adds a delay as it waits for
network responses.

--
Give a man a match, and he'll be warm for a minute, but set him on
fire, and he'll be warm for the rest of his life.

Advanced SPAM filtering software: http://spamassassin.org

From: "Adam Bown" <lists@it-squared.net>

> Can anyone help with some basic stuff...
>
> I have SA set up, spamd is running and exim (with exiscan) runs all
messages
> through it. It's working fine and is filtering messages. However, it
seems
> very "slack". It is filtering REALLY spammy messages ok but 90% of stuff
> get's through. How can I tweak the settings? Is spamd configurable?

Adam, until the Bayesian filter is trained you'll get this. I built up a
nice spam database of over 200 spams within a day or so. Once I had this
and a similar ham database I fed both through salearn in two different
runs. Then I turned on the Bayes filter and spam went away with about
99.8% to 100% efficiency. Training was a bit of a pain. But once trained
the system's pretty darned good.

{^_^}

Thanks for the reply, it sounds like that is my problem. I've found
sa-learn and that looks easy enough to use (although I don't have much spam
to train it with yet so that will have to wait). In the meantime, how do I
check if it's doing the bayes filtering? ...and if not turn it on?

Apologies if this is a dense question but I'm not aware of any configuration
file. (and I confused the issue by installing another "spamd" from the
freebsd ports collection so I'm not 100% sure which files belong to that
one!)

Adam.

-----Original Message-----
From: jdow [mailto:jdow@earthlink.net]
Sent: 14 March 2004 22:12
To: spamassassin-users@incubator.apache.org
Subject: Re: newbie...

From: "Adam Bown" <lists@it-squared.net>

> Can anyone help with some basic stuff...
>
> I have SA set up, spamd is running and exim (with exiscan) runs all
messages
> through it. It's working fine and is filtering messages. However, it
seems
> very "slack". It is filtering REALLY spammy messages ok but 90% of stuff
> get's through. How can I tweak the settings? Is spamd configurable?

Adam, until the Bayesian filter is trained you'll get this. I built up a
nice spam database of over 200 spams within a day or so. Once I had this
and a similar ham database I fed both through salearn in two different
runs. Then I turned on the Bayes filter and spam went away with about
99.8% to 100% efficiency. Training was a bit of a pain. But once trained
the system's pretty darned good.

{^_^}

Thanks for the reply, this kind of brings me onto another question...

Is spamd generally quite slow? If I disable spam scanning in my ACL message
leave my outbox to fast to measure (as you'd expect over a 100Mb network),
but with SA turned on this is more like 3 seconds. Obviously this is no
biggy for light load, but 3 secs per message (plain text without attachments
at that) seems a tad inefficient.

The machine is under very little load so I'm a little puzzled as to why it
might be so slow. Does SA log anywhere?

Adam.

-----Original Message-----
From: Matthew Cline [mailto:matt@nightrealms.com]
Sent: 14 March 2004 14:22
To: spamassassin-users@incubator.apache.org
Subject: Re: newbie...

On Sunday 14 March 2004 10:00 pm, Adam Bown wrote:
> Can anyone help with some basic stuff...
>
> I have SA set up, spamd is running and exim (with exiscan) runs all
> messages through it. It's working fine and is filtering messages.
> However, it seems very "slack". It is filtering REALLY spammy messages ok
> but 90% of stuff get's through. How can I tweak the settings? Is spamd
> configurable?

Enabling network tests can also help, but it slows things down.

--
Give a man a match, and he'll be warm for a minute, but set him on
fire, and he'll be warm for the rest of his life.

Advanced SPAM filtering software: http://spamassassin.org

Hi Adam

I've found spamd to be pretty quick. Our box running spamd probably has
around 3 connections to spamd per second and I would estimate the avg
time to scan per email is ~0.3 seconds (I have set spamc to only send
thru mails < 100k - for our environment I did find the default 250k
limit to high).

Regards
Deon.

On Mon, 2004-03-15 at 01:07, Adam Bown wrote:
> Thanks for the reply, this kind of brings me onto another question...
>
> Is spamd generally quite slow? If I disable spam scanning in my ACL message
> leave my outbox to fast to measure (as you'd expect over a 100Mb network),
> but with SA turned on this is more like 3 seconds. Obviously this is no
> biggy for light load, but 3 secs per message (plain text without attachments
> at that) seems a tad inefficient.
>
> The machine is under very little load so I'm a little puzzled as to why it
> might be so slow. Does SA log anywhere?
>
> Adam.
>
> -----Original Message-----
> From: Matthew Cline [mailto:matt@nightrealms.com]
> Sent: 14 March 2004 14:22
> To: spamassassin-users@incubator.apache.org
> Subject: Re: newbie...
>
> On Sunday 14 March 2004 10:00 pm, Adam Bown wrote:
> > Can anyone help with some basic stuff...
> >
> > I have SA set up, spamd is running and exim (with exiscan) runs all
> > messages through it. It's working fine and is filtering messages.
> > However, it seems very "slack". It is filtering REALLY spammy messages ok
> > but 90% of stuff get's through. How can I tweak the settings? Is spamd
> > configurable?
>
> Enabling network tests can also help, but it slows things down.