BTW, http://plg.uwaterloo.ca/~gvcormac/spamcormack.html seems to indicate that bogofilter is only substantially better by itself if it's trained well. I'ill be curious about combining them, however. That may, unfortunately, involve having to use procmail... --- Dale Kiefling <http://www.outformations.com/~dek> wrote: > Thanks for your time and the notes. > > Please update my email in your address book to http://www.skywriterhosting.com/~dale > > Ciao, > Dale > > On Aug 19, 2007, at 4:56 PM, wrote: > > > 1) bogofilter vs. spam bayes > > 2) spamassassin logging; amavis hiding > > > > 3) try enable pyzor and razor > > 4) enable dcc; use dccproc binary rather than /usr/bin/dcc or not? > > > > 5) bogofilter has multi-token match, i.e., bigrams > > 6) can both spam bayes and bogofilter be used together > > 7) corpus of spam/ham or not?