Paolo <oopla@...> writes:
> On Sat, Dec 16, 2006 at 03:42:55PM -0500, Lloyd Zusman wrote:
>> Therefore, I'm not sure from reading here if one of the other
>> classifiers might work even better with Blame Dalkey than osb unique
>> microgroom. If so, which one might it be?
> try ;)
> seriously, depends on your system and msg+spam mix. IMO it's worth testing
> all but the once-default markovian and the latest bit-entropy classifiers,
> as the 1st one take lot of HDD and CPU more that osb with no proven gain,
> while the latter isn'r stable yet (actually still WIP) and anyway not
> really suitable for text. And at present is by far the more resources-hungry
> But if you have the usual spam stream that plague ~90% of emailboxes, I'd
> say your choices are just osb, osbf, hyperspace, since all reports so far
> show best-of-breed is among these 3. You still have classifier opts to play
> with, but you may want to always keep <microgroom> where supported.
> The 'bad but good' news is that you may need to wait to process a lot of msgs
> to see differences, as all proved to be quite good :).
Yes, this does help.
How can I tell where microgroom is supported? Are the entries in the
"Classfier Flags" section of mailfilter.cf exhaustive (in other words,
can I only use microgroom with osb, osb unique, osbf, and hyperspace)?
Thank you very much.
God bless you.