On Sun, Dec 16, 2007 at 01:14:13AM +1030, Tim wrote:
> On Sat, 2007-12-15 at 08:27 -0600, Aaron Konstam wrote:
> > But the primary emphasis to make SA better is a bayes based learning
> > proceedure. You tell the system about its mistakes and it learns.
>
> Unfortunately, the way most users get to play with that, using the /is/
> and /isn't/ spam buttons on their mail client, is a mystery how it'll do
> the assessing. I don't think I've seen any declare what information is
> programmed into your anti-spam system when you use them.
>
> I wouldn't want to mark several messages that came through a mailing
> list as being spam, and have it decide the mailing list is probably
> spam, rather than the type of information contained in the message.
I dunno about SA, I don't use it. But I use SpamBayes as a filter invoked
from my .procmailrc and it does a fine job. Not more than once or twice
a week some spam lands in a non-spam folder, and less often than that
I get a false positive. That's out of probably at least a couple hundred
spams a day and several hundred mails for various mailing lists.
>
> Unless I know how a mail client is going to use its is/isn't spam
> buttons, I don't use them. There should be some way for me to configure
> my client to take my pressing of the is/isn't spam buttons to mean that
> the sender should be blocked, or addressing should be ignored and other
> criteria be made use of.
>
> Evolution, for instance, gives no clues in the documentation for what
> actually happens when you use its junk mail features. Another of those
> don't look at that man behind the curtain situations.
>
--
---- Fred Smith -- fredex@xxxxxxxxxxxxxxxxxxxxxx ------------------------ ----
Do you not know? Have you not heard?
The LORD is the everlasting God, the Creator of the ends of the earth.
He will not grow tired or weary, and his understanding no one can fathom.
----------------------------- Isaiah 40:28 (niv) -----------------------------
Attachment:
pgpKqEYzQDWrZ.pgp
Description: PGP signature
