[Spambayes] Inconsistencies in classification between my mail reader and the web interface - being flooded by spam

Claude Jones claudejones at tehogeeservices.com
Fri Sep 29 15:24:34 CEST 2006


On Wednesday 27 September 2006 22:10, Claude Jones wrote:
>
> Sometime in the past year, someone incorporated code in Kmail
> that allows it to detect the presence of Spambayes on the
> machine in a wizard interface, and it configures a few filters
> for you if you choose to allow that. The filters are fairly
> innocuous, and seem to do what they're supposed to. But, I'm
> wondering if they're the problem. The filter merely examines
> the spambayes classification line in the header and sends spam
> to the trash bin, and unsure to a folder of your choosing,
> seems right to me. Then, it configures a couple of buttons that
> appear on the toolbar, and those are linked to manual filtering
> actions: "sb_filter.py -s" for the spam button and
> "sb_filter.py -g" for the not spam button.
>
> If the Spambayes proxy-server is processing the message, how
> could one version appear in my mail reader's viewing pane, and
> a different one appear in the web interface??? This is very
> strange.
>

I think I've found the problem, though the cause is yet not clear, 
so I post this in case others encounter this issue. Kmail has a 
wizard for configuring anti-spam filtering. If you click on it, 
the first thing it does is look for several known anti-spam 
programs running on your system. If it finds any, it then 
produces a list. Highlighting any listed program activates a 
routine in which it then asks where you want to direct detected 
spam, and probable spam in some cases, and a final click writes a 
few rules into the filter stack. One of those rules is to pipe 
all messages through Spambayes to be classified. 

However, on my system as configured, Spambayes is started up at 
boot time, and is configured to run checks on all the mail as it 
comes in by piping it through the proxy server. So, I do a mail 
check; incoming mail goes through the proxy server and then gets 
processed by Kmail; Kmail compares the mail against the filter 
stack, sending known quantities to various folders, and, if the 
mail message hasn't been processed by any filters, one of the 
last filters is a pipe to run a Spambayes check. That's the 
source of the reason one message gets processed, and can appear 
in two different places with different Spambayes header 
classifications. It's literally getting two Spambayes checks. 

In my cited example, the incoming message went through the proxy 
server and was classified as spam. Then, it was processed by 
Kmail, and didn't meet any of the filter criteria so it was then 
piped through a Spambayes check for A SECOND TIME. Now, the only 
thing that is confusing to me is why there was one  analysis that 
concluded it was spam, and a second that concluded it was 
unsure... 

I may be on to the answer, partially. Each time I did this, I was 
starting over my database anew by deleting three files, which was 
not quite what the faq instructions stated, by the way - they 
were .hammie.db hammie.db and spambayes.messageinfo.db
Note the two versions of hammie.db - I believe the one without the 
dot at the beginnning was set up the Kmail wizard - that would 
mean that the incoming messages were being processed against two 
different databases! - I guess that would cause the behavior I 
experienced. I note that now, I've removed the pipe in my Kmail 
filter stack that sends the message through Spambayes a second 
time, and now,  no hammie.db file has been created, only 
the .hammie.db ---

Well, I had no responses to my first post, but, maybe this will be 
helpful to someone. If I don't post further followup, assume that 
the issue is solved. I only had one false negative in my inbox 
this morning, and about 11 unsures. That's a huge improvement 
over what I had been getting, just after one evening's training 
of a new database. 
-- 
Claude Jones
Brunswick, MD, USA


More information about the SpamBayes mailing list