Re: spam, imap: false pos: use-BBDB-exclusive, use-blacklist,

[Uwe Brauer]

>>>>> "Ted" == Ted Zlatanov <tzz@lifelogs.com> writes:


    Ted> Open the article(s) in question, hit `t' to display
    Ted> everything, then run the following functions:

    Ted> M-: (spam-split 'spam-use-BBDB 'spam-use-blacklist
    Ted> "SPAM.REAL") M-: (spam-split 'spam-use-BBDB-exclusive
    Ted> "SPAM.HAM")

    Ted> What do you get?

    Ted> These two tests simulate exactly what happens when the mail
    Ted> is split.  If you get the results you expect, something else
    Ted> is wrong.  If you don't, you need to send me your BBDB,
    Ted> blacklist file, and message(s) that are not processed
    Ted> correctly.

Ok now after some testing, there is in fact a problem:
I never receive anything in SPAM.HAM (which is the place for possible
false positive)

More concrete  a couple of mails, which are NOT in my BBDB and NOT in
my blacklist ending up in SPAM.REAL instead of SPAM.HAM
    >> I get
    >>
    >> M-: (spam-split 'spam-use-BBDB 'spam-use-blacklist "SPAM.REAL")
    >> "SPAM.REAL"
    >>
    >> M-: (spam-split 'spam-use-BBDB-exclusive "SPAM.HAM") "SPAM.HAM"

Again, when I apply the above functions I get "SPAM.REAL" *AND*
"SPAM.HAM" that looks wired it should be either "SPAM.REAL" *OR*
(exclusively) "SPAM.HAM".

I send you privately the mail in question together with my blacklist
(my BBDB seems not relevant here.)


Thanks 

Uwe 


    Ted> Generally, you re-train on them, un-registering them as spam
    Ted> and then registering them as ham.

Right, but could I use in principle spam-stat, even on a imap server,
either with agentize or with 
(setq    nnimap-split-download-body   t)?