Spam avoidance

[Douglas Alan]

I've noticed that there is little to no spam in comp.lang.python and
am wondering how this is accomplished. Is there a moderator who
actively cancels spam? If so, that wouldn't seem to prevent spam from
making it through to the mailing list version of the newsgroup. Is
there an exceptionally good spam filter in place? If so, I haven't
previously seen one that works so well. Have spammers just given up
on Usenet, figuring that there are more exciting places to spam than
the Big 8 dinosaur?

The reason that I ask is that I help to run another newsgroup
(rec.music.gaffa) and we avoid spam by requiring that every
participant who wants to post has to register on a Mailman. But I
think that this approach has the seriously detrimental effect of
discouraging new participants, so I've been wondering about
alternative approaches.

|>oug

[skip]

Doug> I've noticed that there is little to no spam in comp.lang.python
Doug> and am wondering how this is accomplished.

Most mailing lists which originate on mail.python.org have SpamBayes
filtering in front of them. The one notable exception are the two
SpamBayes-related mailing lists.

Skip

[Tim Peters]

[Douglas Alan]

I've noticed that there is little to no spam in comp.lang.python
and am wondering how this is accomplished.

[Skip Montanaro] Most mailing lists which originate on mail.python.org have SpamBayes
filtering in front of them.
Worth noting that the SpamBayes project started specifically because
Barry Warsaw tricked me ;-) into thinking about a way to do spam
filitering for GNU Mailman lists. My original "ham" test data was in
fact a year's worth of comp.lang.python postings. The algorithms work
extremely well for a focused tech mailing list (I don't know how they
work on other kinds of lists, because those weren't tested).

BTW, python.org uses other gimmicks too, right? For example, I think
Greg Ward set up some other gimmicks to weed out obvious viruses.
The one notable exception are the two SpamBayes-related mailing
lists.

Yup, and they do get their fair share of spam.

[Terry Reedy]

"Douglas Alan" <ne****@mit.edu> wrote in message
news:m2************@lexx.mit.edu...

I've noticed that there is little to no spam in comp.lang.python and
am wondering how this is accomplished. Is there a moderator who
actively cancels spam?
For the newsgroup, as opposed to the maillist, there is no moderators.
There may be readers who cancel spam.
If so, that wouldn't seem to prevent spam from
making it through to the mailing list version of the newsgroup. Is
there an exceptionally good spam filter in place? If so, I haven't
previously seen one that works so well. Have spammers just given up
on Usenet, figuring that there are more exciting places to spam than
the Big 8 dinosaur?
People who spam the 'dinosaur' groups tend to lose their newsgroup
accounts. News sites that cater to spammers and refuse to cancel can be
blacklisted and disconnected from the main body of Usenet. (I know this
was done in the 90s, and presume threat still exists.)
The reason that I ask is that I help to run another newsgroup
(rec.music.gaffa) and we avoid spam by requiring that every
participant who wants to post has to register on a Mailman. But I
think that this approach has the seriously detrimental effect of
discouraging new participants, so I've been wondering about
alternative approaches.

news.gmane.net, which turns technical mailing lists into free-access
newsgroups, sends a note to first time posters (per list, I believe) and
requires a response before sending the message onward. Was no problem for
me, as I appreciate the filtering.

Terry Jan Reedy

[Neil Schemenauer]

Tim Peters <ti********@gmail.com> wrote:

[Douglas Alan]

I've noticed that there is little to no spam in comp.lang.python
and am wondering how this is accomplished.

[Skip Montanaro]

Most mailing lists which originate on mail.python.org have SpamBayes
filtering in front of them.

BTW, python.org uses other gimmicks too, right? For example, I think
Greg Ward set up some other gimmicks to weed out obvious viruses.

I'm mostly the guilty party at the moment. Incoming mail on
mail.python.org goes through an SMTP server implemented in Python.
The server uses SpamBayes to filter spam. We disallow attachments
with executable filenames (e.g. .scr). That kills almost all virus
mail. We use a number of realtime blackhole lists; they also block
quite a lot of virus junk and some spam. There is a set of manually
maintained message patterns; those kill some annoying junk that's
hard to block in other ways. We do greylisting (two different
kinds, actually). Some IP addresses get blackholed using iptables
(e.g. zombie machines blasting out virus junk). If SpamBayes is
unsure about a message to a list then it gets held for moderation.

I suspect there are people working behind the scenes to cleanup the
NNTP feed. The short answer to Douglas's question: good tools and a
fair amount of elbow grease. :-)

Neil