response 100 of 480:

Nov 28 18:57 UTC 2006

It would certainly be convenient for Grex users if there was a way to simply
select a Grex spam filter that is maintained for the deluge of current spam.
Having every user do it individually is a huge waste of users' time, and
many may not know how. CAEN maintains a Brightmail spam list. I think this
costs them money, but a simpler one could surely be maintained by a Grex
volunteer. 

response 102 of 480:

Nov 28 22:08 UTC 2006

The only people I know (...know of...) that could do it are here on Grex.
Please would one of YOU implement a system-wide spam filter? 

response 104 of 480:

Nov 28 22:53 UTC 2006

I wrote a simple filter that anyone who knows how to copy can copy, but there
are people who don't know anything but bbs shell.  And other people who really
want to get all their spam, including 60 a day from debora*.  The stock spams
have slipped through my filter again, sort of like malaria mutating every day.
Today I am filtering on Target:  Some days it is Projected price: or
Projected: .  They seem to be from .ua or .ru or .mx - should grex reject all
Mexican ISPs?

response 106 of 480:

Nov 29 01:22 UTC 2006

 re #98:  picospan has had the "scribble" segv at least since we moved
 to OpenBSD.

response 108 of 480:

Nov 29 03:54 UTC 2006

What's really necessary is for some staffer to make this their pet
project and bang something together.  Dan seems like as good a choice
for that as any.  Going after the problem piecemeal is not going to be
nearly as effective.

response 110 of 480:

Nov 29 08:45 UTC 2006

I have two silly questions regarding filtering spam: 

 - Is there a tutorial or primer on setting up one's account to use
spamc or some other intelligent mail-cleaning bot? 
 - Do the mail-cleaning bots use a shared (system-wide) corpus of
mail-examples to learn from, or does it have to learn seperately for
each user? 

response 112 of 480:

Nov 29 14:05 UTC 2006

Re #110: I posted an item a while back on how users can set up spamc. 
It's in last summer's Agora conference.  See item:oldagara,236.  For 
convenience, I've copied the item text to the file ~remmers/www/
spamc.txt.
You can also read it on the web at http://grex.org/~remmers/spamc.txt .
The process is pretty simple and involves creating a couple of files.  
You can configure how "aggressive" you want the spam filtering to be.

I think that a reasonable way to make this approach accessible to more 
people (who don't necessarily read agora or even know how to edit files) 
would be to have a standalone program that a user could run to create 
the necessary files and specify the level of spam filtering wanted.  The 
existence of the program could be mentioned prominently in the motd.  It 
could also be made accessible via a "menu" option, or even a web form.

Enabling spamc in your account means that every mail message you receive 
is first filtered through SpamAssassin.  Some staff members have 
expressed concern that this could swamp Grex's CPU if too many users do 
this.  Although this may be true, spam has become so annoying to so many 
users that my inclination would be to try it anyway (unless somebody 
comes up with a better solution Real Soon Now), but make it "opt-in" by 
requiring users to run the program, rather than making spam filtering an 
automatic default for all users.

In addition to the concern about system load, there's also the question 
of what to do with spam messages.  Discard them?  Save them to a 
separate "junk" folder in the user's directory, so that the user can 
examine them for false positives?  The latter approach could eat up disk 
space real quick.

If people think this is a reasonable approach, I'm willing to work on 
implementing it.

response 114 of 480:

Nov 29 14:14 UTC 2006

I'll add that another idea discussed at last night's board meeting was to 
offer users the option of turning off external mail, i.e. only receive 
mail from other Grex users.  Once implemented, presumably that choice 
could also be built into a general-purpose mail configuration program that 
users could run.

response 116 of 480:

Nov 29 15:07 UTC 2006

Thank you for the instructions. While I have no problem with the
instructions given (they look very straight forward), I can see how they
would give someone who has never really worked in a UNIX/BSD environment
fits (I remember how alien pipes and regexes were to me all those many
years ago). 

I applaud your offer to make a script or tool to allow users to easily
set this up for themselves, and if I make it out to AnnArbor, I'll buy
you a beer for your effort. It does not look like a horribly nasty
script, but I would not trust myself with other folk's emails (I know
how pissed off I would be if a well-intentioned but badly executed
script from a college kid blew away a letter from a long-lost friend),
so I am not stepping up to write it myself. 

response 118 of 480:

Nov 29 15:16 UTC 2006

Just wondering, is there a mechanism by which one could mark a message
as spam if it gets through the filter so that the filter's engines learn
from it? I know spamassassin is supposed to be adaptive and adaptable.
How can one help train the system? 

response 120 of 480:

Nov 29 15:52 UTC 2006

Re #117: Not the "maus" you think it is, I suspect.

response 122 of 480:

Nov 29 16:54 UTC 2006

I wrote up a small easy filter that you can just copy from my home directory
to yours, along with my .forward file.  cp ~keesan/procmail.simple
./.procmailrc.  Then change 'keesan' to your own login, and change the
'jdeigert' in my whitelist to the name of someone you want to get mail from.
This filters on anything assigned five points by spamassassin but I would
change it to three points (*/*/* instead of */*/*/*/*) because I never got
a false positive that way.   Someone else copied this but did not let me know
yet if it worked.  A slightly more complicated sample is procmail.sample .
I think I set this to send */*/* to /dev/null and */* to a spam folder.
Today I got no spam in any folder, after adding a few more filters on such
things as Windows character sets, embedded images, From: debora .
I don't recall if my sample filter keeps a log of what went where, but I have
my own filter set up to keep a short version, which is running 20 pages a day
of mostly spam (at 3 lines per entry).

response 124 of 480:

Nov 29 22:41 UTC 2006

Using the method I described, it's easy to incorporate that.