Beta 4.50.4 released -- faster than 4.49

[Randal, Phil]

Raymond Dijkxhoorn wrote:
> Its in my eyes again comming to the same basic point, low
> spam you want to do different things with, so the score can
> reach the high value in some time. Bayes is one of them...

How we should behave could depend on our high/low spam score actions,
too.

If we quarantine but don't deliver low score spam, then it doesn't
matter so much that it isn't having its score ramped up by Bayes
learning, except that Bayes learning could help push future variations
of that spam from ham to spam scores.

Which leads me to think that while spamassassin is flagging mails as
"autolearn=spam" they shouldn't be cached.  Though if they are
high-scoring already we could still cache them without materially
affecting the Bayesian stuff.

The other thing which occurs to me as highly desirable is a "service
MailScanner flushcache" parameter, and possibly a "flush cache on
restart" option in MailScanner.conf, so we can flush the cache after
changing Spamassassin rules (either manually or automatically in a
RulesDuJour type script).

The last thing we want is to create a broken rule which flags mail from
a legitimate mailing list as spam, discover our error, correct it, and
still have straggling emails from that mailing list (to different
recipients) being flagged erroneously as spam.

Aren't cache coherency issues fun?

Cheers,

Phil 
----
Phil Randal
Network Engineer
Herefordshire Council
Hereford, UK

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the Wiki (http://wiki.mailscanner.info/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!