On Sun, 07 May 2017 15:05:21 +0100 Martin Gregorie wrote: > On Sun, 2017-05-07 at 13:27 +0100, RW wrote: > > On Sun, 07 May 2017 10:33:56 +0100 > > Martin Gregorie wrote: > > The only other thing you need is a way remove SA headers from your > > > spam collection. A bash script using awk to do the heavy lifting > > > works for me: > > > > > > > That shouldn't be necessary. AFAIK as all the X-Spam-* headers are > > stripped before Bayes can see them. The common SA headers are also > > in the default "bayes_ignore_header". > > > Fair enough for playing with Bayes, but leaving them in place is a > just nuisance when you're developing private rules because the old > headers tend to obscure what you're looking for.
Actually I find it practically essential to leave them in. For example if you want to see all spams that hit both pyzor and razor and scored under 12, how else are you going to do it? > If you push a message through spamc with X-Spam* headers already in > it, the original headers are still there in the output, though now > preceded by the ones added by the test. I've never seen that happen because, as I said, SA strips them. It may happen if some third-party software writes the headers.
