On Thu, 2003-08-28 at 11:50, Carlo Wood wrote:
> Are you 100% sure that is also the case for RBL checks?
> It seems 'weird' that this test will indeed be completely
> turned off if, and only if, all six related scores are set
> to 0 (and not when you forget one). That seems like an
> almost complicated thing to code and will have to be
> specifically taken into consideration when writing that
> code in order to work. Was that really done?
Its not that hard to code... I've not actually read the code for this
part of SA, but if I was doing this I would cache all network tests
aggressively - if there are 2 rules looking for different values for the
same key in an RBL I *really* don't want 2 different lookups to be done
- network operations suck for speed, and memory is a lot cheaper than
that sort of latency.
So I would process a (RBL type rule) as:-
* Look at the score - if zero, skip to next test
* Call RBL lookup/cache with appropriate key
* check value against trigger value, if a hit then increment
running score
That would work quite nicely for both minimising lookups and skipping
doing the lookup if there was a zero score.
Nigel.
--
[ Nigel Metheringham [EMAIL PROTECTED] ]
[ - Comments in this message are my own and not ITO opinion/policy - ]
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
