Matt Hampton writes: > Evening > > I have been running the updating rules for just under two weeks. The > results are fairly good very few FP's and the ones I have been seeing > are on ones that were meant to be purely triggers for META rules - will > have to introduce another set which demotes them back to non scoring rules. > > I would be interested in hearing what other people get so that I can > tweak my selection process - the latest one has ~140 rules in - I think > this low number is more an affect of my corpus size than anything else - > mass-checks is running on an older box which grinds everything else to a > halt when its running so haven't got a huge number on there. > > I have created a sa-update channel - but won't publish the domain at > present until I have spoken to Sane again - haven't heard from Steve yet > - am going to try again on Monday. I am automatically generating them > but manually publishing them at present - the latest published file is > avaliable from http://www.coders.co.uk/80_sane.cf
hi Matt -- I've copied yesterday's file into SpamAssassin SVN for testing -- keep an eye on http://ruleqa.spamassassin.org/ , results on our test corpora should appear tomorrow. > As a side issue - I am running sa-compile on my production boxes and I > have examined the .re files generated and can find neither mine or JM's > SOUGHT rules ending up in there. > > Is this expected or is there something about my rules that don't meet > the criteria for the extraction process? I am using body rules.... No, they should be getting in there. You could try running sa-compile with "-D" and ensure it's reading the right rule files... --j.
