Hi all --
If anyone's been checking taint.org, they might have noticed that I'm back
in contact -- although I'm taking my sweet time to get my mail setup
working decently again ;)
Anyway, some quick notes on SpamAssassin dev based on a skim of recent
traffic:
- regarding lists and 'my address' stuff. One thing I should note is that
the *previous* spam filter I used supported both recognising "legit"
list mails, and recognising your own address in order to penalise mails
which aren't explicitly addressed to you. Here's why they were bad IMO:
- spammers are already using "legit" list manager software, possibly to
get past some filters - I've seen Lyris ListManager used (iirc).
Probably helps in bulk-mailing large lists of recipients, too, so it
makes sense.
- Dunno about you guys, but I get a whole load of spam via mailing
lists whose addresses are on CDs. I want SpamAssassin to catch spam sent
to a list, as well as it catches spam sent to me directly.
- Using knowledge of "the user's address" to figure out if a mail was
sent to you, or to a list, means additional customisation is *required*
for anyone setting up SpamAssassin, unless we figure it out
automatically.
Also means that any forwarding set up from [EMAIL PROTECTED] to
[EMAIL PROTECTED], will require modification of the "my address" setting.
Ditto for any new mailing lists you may subscribe to. Lots of
customisation required, continually, which is a pain.
Bcc'd mails are already penalised. This will penalise them more.
Now, I can see the top reason it'd be useful, however. Spammers are
using the dest. address as the forged From address, since that is almost
always in the AWL as a non-spamming addr. If we can come up with some
way to work around this *without* requiring a "my address" or "my
domain(s)" setting or regexp -- ie. using some stats analysis on the AWL
data instead -- it would be infinitely preferable...
- I'm going through the CVS version, fixing some over-aggressive or borken
rules. hope no-one minds ;)
- Quick query on a comment in PerMsgStatus: it claims that the body-text
rules are faster if called procedurally. How is this? And how come
passing $_ as an arg instead of \$_ is faster?
I'm sure there'll be more questions, and I'll probably accidentally undo
someone else's rule changes from the past few months, through ignorance of
what's been going on, so I'll apologise in advance ;)
cheers,
--j.
--
'Justin Mason' => { url => 'http://jmason.org/', blog => 'http://taint.org/' }
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
No, I will not fix your computer.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
