On Thu, Oct 23, 2003 at 09:31:50AM -0700, Tom said > On Fri, Oct 24, 2003 at 02:21:41AM +1000, Rob Weir wrote: > > Yup. Reason #23131 why CR is a poor solution. I'm quite amazed at how > > well SA and simple checks like my Postfix body regexp work. I still get > > spam, and oodles of it, but it's almost flawlessly classified. I check > > out my spam folders every few days, and I've found perhaps five false > > positive in the past...3 months, more? False negatives are a smallish > > issue, but almost everything that gets through is either too small for > > bayesian to work effectively, or CJK spam that I can't even read. > > I'm on a "the perfect is the enemy of the good" kick these days, so I > understand your happiness with what SpamAssassin is doing. But I guess > spam bothers me so deeply that I feel ONE distraction is equivalent to > ONE THOUSAND distractions -- I still have to peform the mental > operations: (1) Check Mail, (2) Scan Mail for distractions, (3) Delete > distractions. (3) is not the primary pain point for me; (2) is. This is a good point, but it's not something I notice anymore. I scan through my lists and hit "y" on any spam in mutt; it passes the mail to "sa-learn --spam" and moves it to my spam folder. About the only thing I see anymore in the Debian lists are CJK spam, so subjects like ???????????????????????????????????? activate my "y"-reflex. I'm considering some further measure to bind to my magic "y"-key. Reporting to spamcop was high on the list, but I've heard rumours of them accepting payoffs in exchange for ignoring some spamhauses, so we shall see. Bouncing the mail to postmaster@last_hop_before_me is something I'm a little nervous of the consequences. Perhaps spooling a weeks worth of spam for each relay then sending in a bunch is better. I'd have to be *extremely* careful not to bother listmasters with this sort of strategy, of course. It seems that blocking at the MTA level is becoming the only real option to handle the flood of crap on the internet. > So I'm on a quest for zero-spam solutions. My current idea is > disposable email addys; if I can go ten days on a list mail or two > months for a friends/family email, I'm okay with that. I track who has > what and send out update notices (or rejoin lists) when I change; note > the date encoded in my addy. I haven't scripted it yet, but I could. I do something similar right now. Everytime "something" I don't trust asks for an email address, I make up a new one: nyt-register-2003-10-23@foobar.ertius.org, for example. Thus, any spam I get from that source will be clearly marked. If they do spam me, I just have to reconfigure my MTA to reject that address. -- Rob Weir <rweir@ertius.org> | mlspam@ertius.org | Do I look like I want a CC? Words of the day: SWAT ASO munitions RSA rs9512c radar security Marxist
Attachment:
signature.asc
Description: Digital signature