=== New Antispam ===
Today is a special edition about mail. From September 2007 we've been
using DSPAM, a very nice piece of antispam technology. Instead of
neverending rules fiddling this system properly learns to recognize
unwanted mails. Unfortunately it has been more than four years that
maintenance stopped and we've been having a hard time finding a
replacement. We believe we did and now is time to give a chance to the
new system to prove it. Thanks a lot to all the DSPAM contributors, that
was really awesome.
Our new system is Rspamd. The new system uses a mix of solutions to
score mails, including headers analysis, Bayesian rules, adaptative
greylisting (activates if mail appears somewhat spammy), RBLs… We've
selected rules that would not leak mail content onto third parties.
We've also decided to drop per-user learning, as it really makes the
architecture more complex and consume more resources (early rejection
would no more be possible), but current tests shows it should work out
Currently Rspamd and DSPAM are working concurrently (for filtering and
learning) but we intend to stop DSPAM very soon. Mail filtering should
continue to work as usual. I updated the documentation accordingly:
Outgoing mails are now checked by Rdpamd too.
Please let us know if you have any problem. The system also needs more
input to learn but it should not take long. Don't forget to move in/out
the `Junk` folder SPAM/HAM mails to teach the system.
=== Some words about the future ===
We're currently transforming Toushirou into a full mail server (MX1), so
when Orfeo is in maintenance or a problem occurs, we should be able to
read our mails and receive new ones.
We needed more storage for Toushirou-NG and Pilou offered new disks,
thanks a lot :-). We're working on bringing Toushirou's new body to
Have a Lot of FUN!