On Wed, 17 Mar 2004 16:31:18 -0600 (CST), Adam D. Lopresto wrote:
> On Wed, 17 Mar 2004, Mat Bowen wrote:
>
>> What about only running the RBLs if the email is below the spam
>> threshold? Most of my mail is classified as spam without running
>> them
>> so it seems unnecessary to spend time checking them only to push
>> the
>> score up higher.
>
> The RBLs take a lot of time in wallclock terms, but not much
> processor time.
> I could see the use of firing off the RBLs at the beginning of the
> run, though,
> and when you finish with the regexps, if you're already over the
> threshold
> don't bother waiting for the RBLs to come back.

I can't imagine the non-RBL tests take much wallclock time to run
compared to the wallclock time of the RBL tests (maybe i'm incorrect
in thinking that though) so the time saved from running the RBL tests
first would be minimal. Running them conditionally at the end of all
other tests (if the spam threshold hasn't been reached) has the
advantage of potentially saving a lot of strain on the servers
involved as well as the advantage of saving the wallclock time on the
local machine if the tests aren't needed. Or perhaps it could be an
option to do it either way :-)

Mat


Reply via email to