DSPAM v3.2 Beta-1 Released

Nuclear Elephant writes "After three months of development, the first public beta of DSPAM v3.2 has been released for testing. New features include SQLite support, A Win32 build supplement, extensions API, and some advanced new processing functionality such as Bill Yerazunis' (CRM114) Sparse Binary Polynomial Hashing and v1.2 of the author's Bayesian Noise Reduction Logic. Accuracy in 3.x has reportedly peaked as high as 99.991% (2 errors in 22,786 messages). Grab the new copy and participate in the request for feedback."

Comments

Re:SpamAssassin vs. DSPAM

[jwbozzy]

I totally agree. I used DSPAM for a while, gave it a fair shot, participated on the mailing list. I even, at times, got encouraging results. Ultimately, DSPAM required way too much nursemaid work to make it work for my installation and I scrapped it and went back to SA. The general feel I got from the DSPAM crowd was a big dick waving contest with other products, particularly, but not limited to, SA. A typical mailing list message looked like:

"SA is able to do X accurately. I cannot seem to achieve this with DSPAM. Am I doing something wrong, is there something I need to configure further?"
"SA is inferior. You don't want X. Besides, DSPAM has Y, which approximates X. No, I can't tell you how to do it specifically, but know that you need only DSPAM."

Personally, I found that DSPAM is blatently unable to train itself properly. You might have to train it 4 or 5 times with the same message to get it to classify that message as spam. It doesn't recursively train like SA does. This leads to users getting the EXACT SAME spam multiple times, despite their best efforts to train the filter. In addition, DSPAM's group features are sparsely documented and somewhat magical in their behavior. And of course, with out these features, DSPAM is useless to an installation of people who really do NOT want to have train their spam filter for months to get it to work right.

I'm all for competing software/products, but both projects are OSS, there is no money involved here, and I can't see how bashing the other product while concurrently not being able to do better, or even match it can be viewed as a step forward...

But hey, that's just me, your mileage may vary.

DSPAM. . . neat at fist, not for long.

[Christopher Cashell]

I used DSPAM for a while. I started using it with the Berkeley DB backend, and that worked reasonably well. . . it was fairly fast, but database corruption was almost impossible to avoid. I don't think I ever managed more than 3-4 weeks without my DB getting killed.

So, then I started using an SQL database. That worked great for a while, except it was slow. Now, admittedly, I'm running my mail server on an old machine (Dual Pentium Pro 200's, with 450MB RAM), but DSPAM was horrible. With more than half a dozen e-mails to process at a time, it would just choke. And the space issue. . . my spam-data database got over 300MB within a couple of weeks! And, yeah, I was processing a lot of mail, but come on. That's just not right.

Finally, I decided it just wasn't worth it. So, I tried an alternative that the DSPAM author has spoken fairly highly of, CRM114. That thing rocks! Within a few days, it was catching most of the spam, it runs much faster than DSPAM or SA, and it has fixed-sized spam token databases, so unless you explicitely increase the size, they won't grow past what you set them up for.

I can't see myself bothering with any other spam filter anytime soon.

Re:DSPAM. . . neat at fist, not for long.

[Anonymous Coward]

Sounds like you weren't purging, and this was probably why you had severe space and processing issues. You can switch to TOE-mode training and make your databases as small and fast as CRM114.