dawa wrote:
Hello!
System: Debian 4.0
SpamAssassin version: 3.2.2
MTA: Postfix with Amavis and ClamAV
- - - - - - - - - - - -
We have the problem that some very obvious spam mail come through (example:
http://pastebin.com/m606b1420).
it now hits URIBL_BLACK.
Here, RCVD_IN_PBL is set to 4.0.
Now I thougt bayes is on (see here the local.cf:
http://pastebin.com/m55ba168) but I think it didn't work correctly.
you disabled Bayes with your:
use_bayes 0
With Bayes on, HTML_EXTRA_CLOSE would get 2.8.
A test with "spamassassin -D --lint" will show this output:
http://pastebin.com/m42478b77
I want to let learn Bayes again but all mails are not get stored on this
mail server but get fetched from a lotus notes server.
I saved about 200 not identified spam-mails in notes so I must copy out the
header from each and copy it as a text file in a folder on the Spamassassin
server and let Bayes learn it again!?
you can use fetchmail or the like. ideally, you should copy "unaltered"
messages (they can have few additionnal headers of course).
but if that's too much, just wait and use "future" mail instead.
And is it ok when after the learning of the spam I activate the bayes
autolearn?!
activate Bayes and autolearn now. There is no reason to wait until you
train it.
What else can I do to identify obvious mails?!
if you got the mail directly (and not from your ISP/MSP/... forwarder)
and depending on your setup, you could reject at smtp time using
zen.spamhaus.org. and even if you don't, I've never seen anything real
from p[A-Z0-9]+.dip0.t-ipconnect.de.
