Re: consistant failures on a new kind of spam

> Yeah that's a very cool idea.  I'll have to think about the implications
> a bit since I would not want to make any remaining terms unreachable.

Thanks, glad you like it Brian! I'd be willing to help test any bleeding 
edge changes as I can put spamprobe in front of some very aggressive spam 
volumes.

> I suppose it could use the standard cleanup parameters (min term count,
> max time since last use, etc) to decide when to overwrite a term during
> a collision.

Yes, with those cleanup parameters in place I don't see how the collision-
driven cleanup could possibly become "any worse" than a standard cleanup. 
For instance if an attacker tried to craft terms to cause hash collisions, 
she could not get past the min term / max time checks.

> Yeah, I like this idea a lot at the moment.

David just sent a message that perfectly demonstrates this cleanup 
confusion which motivated the idea:

> > but I really just guessed these numbers.   I understand why brian
> > doesnt want to give us "the perfect numbers" to use because there
> > are none
> > ...
> > So I'm thinking I should make a bigger DB ... or use 
> > different cleanup numbers ?   How to choose !
> > Most of us dont have the time or even desire to really keep track
> > of this stuff. I'd love some procedure that simple "works prety good
> > without failing badly"

The clean on collision, and some kind of verbose output on iterations 
before insert, would satisfy what he asks - an automatic process 
(collisions) ; reasonable defaults (less sensitive due to new random 
collision factor) ; and feedback guiding user selection of database size.

- Jem Berkes



-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV