> On 21/08/15 18:20, Chris Siebenmann wrote:
> > Our incomplete local solution is that all forwarded email that we
> > think is spam is shuffled off to a separate Exim server
>
> If you can identify items, you can choose an alternate outbound IP
> without needing a second Exim installation.
We started out with that approach but we found that it confused
Exim's retry handling too much (and made for messy queues too). Exim's
retry handling for domains doesn't (or didn't) really have support
for multiple sets of retry timers for different (abstract) types of
email. We wound up doing all sorts of things to, for example, make
sure that old 'probably spam' email would be deleted by router rules
before it reached the normal retry timeout and thus potentially trigger
unwanted rejections for non-spam email.
In practice, running multiple Exim installations with separate queues
has wound up being much simpler. We get to not worry about all sorts of
issues and it also makes monitoring simpler because when we look at eg
queue sizes we don't have to try to sort out how many are probably-spam
(and expected to backlog) and how many aren't (where a backlog may be a
problem we need to look into).
- cks
