Alton Danks wrote:
[...]I suspect some of the folks asking "why is this one getting through?" aren't!
Is there anyone who is not using Bayes?
Much like Mike, I'm increasingly depending on Bayes as my primary defense. I do use the default rules, plus several of the add-on sets primarily for training bayes. I'm using spamassassin to train up bogofilter (bayes only) to see how well it might function on it's own.
Why not?
Although I'm using bayes, I'm aware of a couple of limitations:
1. For a small site, a shared bayes configuration may work. But with larger populations, with widely varying tastes, individual configurations are probably preferable. However, if not trained correctly (requiring user participation), bayes can fail miserably. I can certainly see "admin overload avoidance" as a good reason not to use bayes.
2. Bayes training can go wrong, whereas the fixed rule sets can be added or fixed with no long-term effect. If the bayes database gets hosed, it has to be re-built. Unless you're storing a bunch of normal user mail (with the associated privacy issues), getting it re-trained may take a while.
What are you doing above and beyond the default SA config?
If not using bayes, I would at least be using the network checks.
We are not and I'm wondering just how alone we are.The biggest question is, "is it working for you?" If so, then there may be NO REASON to bother with bayes. However, if enough spam is getting through to be creating a problem, it's another tool in the arsenal.
The power of spamassassin for me is it's multi-layered approach. Flexible pattern matching, network checks plus bayes is particularly potent.
- Bob
