F.A.Q. Tweaking BLAST

Tweaking BLAST

Q Sometimes a message is not rated as spam by blast, but if I recheck, it is!

You are right, and it took me a good while to find that 'bug'. But after that the reason is quite clear:

  1. BLAST rates a message as Non-Spam.
  2. then the DNSBL-detector rates it as spam,
  3. then the BLAST rater learns that this message is spam and
  4. at the end the 'secure message' process changes the content.

Now imagine you do a "this is not spam" and then a "check this message". DNSBL will again identify the message as spam. But BLAST has first unlearned the message and then learned the already secured message.

So the BLAST rater has learned the special HTML-code modifications Disruptor OL does to secure the code of spam messages. And this identifies spam (because normally Disruptor OL only does it with spam).

And the info-about-message-dialog shows info about the secured message (because the unsecured message does not exist any more).

Sum: no bug.

Q How to set up BLAST?

The default BLAST setup is a very conservative approach: it needs more than 1000 messages to start working.

I restarted my own BLAST database multiple times, sometimes I had very good results after 200 messages, sometimes it needed about 300 to 400 messages. In all cases this is not a 'all around careless' situation, it needed some "this is not spam" clicks. After about 600 messages it was almost errorfree for me (about 500 mails per week).

Due to the nature of BLAST these values can not be generalized to every user. (And that is why I choose the conservative default value). But they are a reference if you want BLAST to start earlier.

It is possible to limit the total number of messages in BLAST to less than 4000 (the default). This will speed up the load time/scanning start delay, reduce the time needed to batch the BLAST files and save some disk space.