There are many tactics in fighting spam. Most of them involve e-mail spam. Like many others, i've deployed quite a few tactics in fighting spam. The most recent ones being greylisting, bayesian filter and blacklists.

One natural division for all the spamfighting methods is whether it can be deployed to an ISP environment. I've seen all sorts of methods, others are ones that i would never put in to a production system and others that have no effect in a large scale system. One of these methods is plain and simple strict standards checking, which can very easily go wrong. I agree that this is something that should be required, but since there are too many broken sites out there, there is no place for these kinds of methods in an ISP environment. Same goes for greylisting. People expect mail to arrive immediately, which from a technical standpoint is an insane expectation due to the nature of SMTP as protocol (mails are allowed to be deferred), but it rules out greylisting.

There are middle grounds in all these, whitelist & blacklist + greylist works pretty well, but not quite well enough. But there appears to be a new contender out there. Policyd-weight is one of these, it attempts to make blacklist less strict (sites get blacklisted all the time) so that legimate mail gets through while weighing in the validity of HELO and other fields.

Few days ago i disabled greylisting from my personal site, it rejected so little now that i configured policyd-weight in front of it. Looking at the statistics from my dspam there is no change in the amount of spam hitting it. Too bad i don't have older statistics anywhere, it would have been fun to show numbers here, but i remember that the drop in spam was significant when i first deployed policyd-weight. I'm not saying that greylisting isn't effective, but policyd-weight is more effective (IMHO).

As for the question, is policyd-weight suitable for ISP environment? Yes, i think it is. It's a combination of all methods currently deployed which gracefully handles situations when something is slightly wrong.

I'm a lone spamfighter with a new tool...

Comments on this page are closed.