Aaron Stone wrote:
I've been thinking about supporting the virustest/spamtest Sieve
extensions by way of linking to other popular open source virus and spam
testing libraries. It's a bit of a ways off, and it means tieing things
into Sieve rather than the MTA -> LMTP (...) -> DBMail-LMTP delivery
chain, which if not easier to set up is at least discrete parts that are
individually maintained.

Like many things in the email world, there's a place for both ;-)


My opinion is to keep things highly segregated and simple where you can.
Plug them together to make better applications.

Just like unix pipe (|).

I think it's great that dbmail does one conceptual function (MDA) and does it well. To start adding MTA features to it muddies the picture. Adding spam and virus functions to it also muddies the picture.

With spam and virus checking a lot of people use many different approaches depending on what they feel is important. If dbmail can't address all of these then you have an installation of dbmail that has a lot of features turned off.

An example of these arguments might be:

using clamav on a remote machine that filters all mail -- inbound & outbound for larger companies with many clueless users who will gladly mail I_LOVE_YOU to everything they know. dbmail can't manage outbound virus scanning.

using spam filtering applications that are also on remote machines to handle the additional load or to concentrate all the spam filtering from many mail proxies into one group of spam filtering servers.

using spam filtering as a pre-queue filter instead of a post-queue or pre-delivery. There is an increasing number of applications being developed that sit in front of the MTA to filter for spam. Some instances are postfix with content_filter or standalone applications that monitor port 25. These are different from the content_filter variants for postfix (Amavis) which are considered post_queue. The spam filtering advantage of these is the email is rejected rather than every being accepted by the MTA.

putting spam filtering or virus scanning into dbmail might be able to address some of these. But it can not address the pre-queue or pre-delivery spam filtering mechanisms. So you run the risk of developing and maintaining a block of code that a lot of people may never use may also consider to be bloat.

A personal opinion about amavis and spamassassin is that the number of options that they now present are unwieldy and makes getting the installation correct and logistical nightmare. Sure, you might be able to get things running with only 5 configuration entries -- but there's another 100 to administer if you want to fine tune things. It's a kitchen sink set of applications and one of it's biggest complaints that I hear these days is that it is relatively slow and bloated.

_______________________________________________
DBmail mailing list
DBmail@dbmail.org
https://mailman.fastxs.nl/mailman/listinfo/dbmail

Reply via email to