Hans Sandsdalen wrote:
[EMAIL PROTECTED] skrev:
Tom Smith writes:
Is it safe to assume that most MTAs will give up trying after four
or five days if a message is still undeliverable, even if it's
getting a temporary error? (The spam I referred to is dated in
2003, but I don't want to set a date limit that old--perhaps a week
or two.)
Like I said before, it's sort of safe, but you might also want to
consider rewriting the dates of the e-mails which are sort of strange
but not strange enough for you to block... that way the spam with bad
dates won't go to the top of the users e-mailprograms with a less
than perfect sorting algorithm.
(Something I'm getting a touch of right now as I've had to urgently
move to a solution where my date-rewriter is no longer available and
the mailprogram has a lousy sorting algorithm...)
/Tony
I will suggest using the tests in spamassassin, DATE_IN_PAST_xx_yy.
You can set the score of one of the test high enough that mails are
blocked.
SpamAssassin is pretty good and all, but prone to false-positives (as is
anything that filters email based on content). Spammers are always
coming up with ways to defeat such things and so this endless cycles
continues... Spam software gets betters, spammers defeat it, spam
software adjusts to the new techniques, spammers find a way to beat it,
and so on.
The techniques I employ need to work very well with as minimal a chance
of false-positives as possible. So far, I've achieved this--no valid
email on any of my servers has been blocked in a long time. The other
side of this is that some spam still gets through.
I've tried a number of programs to stop spam, some worked better than
others... But all have their cons. I have yet to find something that I'm
totally satisfied with. (qpsmtpd has worked quite well, I might add.
This is one of the better programs I've implemented in a long time, and
one that I'll continue to use. :-D )
So now I'm reading about how others have implemented spam software like
SpamAssassin, ASK, TMDA, et al. I'm reading about the pros and cons of
such systems and working on a framework to address those problems and
(hopefully) create something that will work exceptionally well and stand
the test of time--that is, something that can't be defeated by spammers.
Do I sound like an idealist? Maybe. But I think some new ideas are in
order and maybe, just maybe, I will bring some of them to the table.
