3.10.1 has been released...
dspam project under new management!
dspam is alive and development continues towards the 3.9.0 release. You can find all the latest info at the dspam website (http://dspam.sourceforge.net/) or over and the sourceforge project pages (http://sourceforge.net/projects/dspam/)
I spent some time on the nuclearelephant site, reading the FAQ and the Readme, as well as exploring.
I had a question that I could not find answered and didn't see any forums, maybe I missed them.
As your FAQ noted, SA is out of date and being circumvented to an alarming degree. My server is a shared server running the Cpanel suite and I was looking for an "installation for dummies" type checklist. (Sorry I am what I am.)
I can FTP the files up and create a MySQL database with users and assign permissions, but I am a little lost at a lot of the other cmd line functions. Does a step by step install for this type of environment exist?
Thanks for your patience.
Re: Can't make heads or tails
> The documentation for dspam is almost
> non-existant. How frustrating. I had to
> make sense of things from the
Hi! Have you checked the doc/ directory, which is full of
documentation for various types of integration, or the very
elaborate README? How about the online FAQ, or the many
links to HOWTOs on the website?
Can't make heads or tails
The documentation for dspam is almost non-existant. How frustrating. I had to make sense of things from the usergroups.
3.1.2 is a recommended upgrade, especially with pgsql
I have upgraded to every 3.1.x release, and am by far the happiest with 3.1.2. They fixed a bunch of stuff, especially concerning the postgresql backend.
My false positive rate has plumetted.
In production, stellar performance
We've been using DSPAM at a small ISP for the past 4 months and it's performing admirably. Not only is DSPAM eerily accurate at eliminating spam while avoiding false positives, but the training process is easy enough for the archetypal "grandma".
Read documentation on permissions very carefully. As with any statistical despamming method, be prepared with decent spam and nonspam corpus material to feed DSPAM on setup if you want instant results.
Using on production server
Dspam is working great for me. I have about 12 users, it works very well for a small set like this. I tried spamassassin also, but found too many exceptions had to be made for each user to avoid false positives.
Dspam is very new, it should get even better. I think the dspam_corpus tool needs some work, it doesn't work as smoothly as you'd think it would. In the docs there needs to be more examples of how to incorporate Dspam into your MTA. Only Sendmail and Exim are shown, and only one example of each. There needs to be many more MTA examples.
Re: Looks promising
I agree - we've been using Spamassassin for about 18 months now, and while it's extremely good, it's not really set up to allow users (non-technical ones) to train their databases very easily. This is mainly becase it requires the spam/ham to be present on the mail server you're training. We use POP3, which effectivly makes this impossible. The DSPAM model of allowing users simply to forward spam to the engine is simplicity itself.
Also, using Spamassassin we've come to realise that unless filters are heavily customised to the individual user, the number of false positives is just too high to use conventional heuristic filters. And setting the spam score higher just means more spam gets through - and usually just the worst kind.
Bayesian is the way to go we think. We're going to have a very good look at DSPAM.
I'm evaluating this product... looks very
promising to me.
All the best, hopefully you will attract some
An open, cross-platform journaling program.
A scientific plotting package.