On Fri, 27 Aug 2010, Bren Letson wrote:

I scanned subject lines in the list archives back six months and didn't see anything that indicated this idea has been proposed before. However, it seems unlikely that I'm the first to have these thoughts:

You're not. The term we use is "shortcut"

  1. When looping through rules, stop the loop on the last rule or,
     optionally, when score_so_far >= required_score. This would
     improve performance, I suspect.

What happens if later rules would reduce the score back below the threshold?

  2. Order rules so that statistically more likely rules with higher
     scores appear first. This in combination with suggestion #1 could
     also improve performance. Of course, it implies that part of
     "learning" is ranking how often specific rules get triggered. This
     ranking might be a function of the DNSBLs of the world.

If there was some way to search the list for this idea, I didn't see it. If these ideas have already been discussed, please forgive this newbie.

It has been discussed. It's a non-trivial problem. If you search on "shortcut" (on the users list, too) you'll be able to find a lot of the history.

--
 John Hardin KA7OHZ                    http://www.impsec.org/~jhardin/
 [email protected]    FALaholic #11174     pgpk -a [email protected]
 key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C  AF76 D822 E6E6 B873 2E79
-----------------------------------------------------------------------
  The first time I saw a bagpipe, I thought the player was torturing
  an octopus. I was amazed they could scream so loudly.
                                        -- cat_herder_5263 on Y! SCOX
-----------------------------------------------------------------------
 Tomorrow: Exercise Your Rights day

Reply via email to