https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7742

            Bug ID: 7742
           Summary: "delay delivery" threshold
           Product: Spamassassin
           Version: 3.4.0
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: spamc/spamd
          Assignee: [email protected]
          Reporter: [email protected]
  Target Milestone: Undefined

So, it seems that the root cause of #7739 was in fact that I seem to be getting
spam slightly before it's sources have hit the BLs.  I know for a fact this was
the case for PSBL in #7739 and suspect as much for spamhaus.org's BLs, but
since their lookup service doesn't provide "date listed", I can only assume.

But that had me to thinking... this must be a general problem and could become
more prevalent if/when spammers are able to amass such distributed resources as
to spam just about everybody within a very short period of time.

A defence against this would be to have a "might be spam" threshold below the
"is spam" threshold, and when an item of e-mail is between those values, SA
puts it into a queue and reruns (perhaps only previously failed) tests on it to
see if it has become more spam-looking than when last tried.

At this point several things could happen.  If the score didn't increase any
more since the last try, release it to deliver.  If it exceeded the "is spam"
threshold, treat as spam per usual.  If it's still between the two thresholds,
hold on to it a little longer and repeat previously failed tests again.  Rinse,
repeat, until it's no longer increasing in value, a (hold) time-limit has been
met (and it's still not spam), and which point release it for delivery.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to