[Mailman-Developers] GSoC 2014 : Proposal for the Mailman CLI project
Stephen J. Turnbull
stephen at xemacs.org
Mon Apr 28 06:35:44 CEST 2014
> On 4/27/2014 11:03 AM, Stephen J. Turnbull <stephen at xemacs.org> wrote:
> > When you get ~250 wanted mails (many of them list, of
> > course) and ~1000 spams (that get past the 6-sigma "if this filter
> > thinks it's spam, throw it away!" filter) a day, automatic processing
> > is really important.
> Anyone who gets ~1000 spams per day that actually make it through
> whatever anti-spam tools you are employing,
I didn't say that I actually see them, I said I get 1000 that can't be
rejected/discarded as spam with 6-sigma accuracy. Dealing with the
1-in-100 Type II errors and the 2-in-100 Type I errors in the 1000 is
what much of the "automatic processing" is for.
> then you need different/better anti-spam tools.
Suggestions are welcome. Most of the problematic mail is in Japanese
or Chinese however, and I don't know any tools (including GMail which
throws up false positives in my spam folder about once a week) that
get 2- or 3-sigma performance on those languages, at least not in my
multilingual context. So I quarantine, and do a lot of tweaking of
packaged tools and postprocessing myself.
At least some commercial tools for Japanese are really horrible --
about once a week I get mail from a colleague that is marked as "spam"
or "suspected spam" by my *employer*'s filters, and traffic on this
list gets marked that way about as often. :-(
More information about the Mailman-Developers