[kwlug-disc] (Running a VPS) Mail Server? [

[Hubert Chathi]

On Fri, 19 Dec 2014 09:49:16 +0000 (UTC), "B. S." <bs27975 at yahoo.ca> said:

> Speaking of ...

> The idea of running one's own mail server has always seemed a little
> 'scary' - not just to me, but others here have expressed similar.

> So many little fiddly bits, DNS stuff [can't recall the term, not DFC
> - the MX record isn't the problem, it's the rest], multiple programs
> chained together, blacklists, spamassasin, don't know what all else.

> Always seemed too much work. Anyone using yahoo these days probably
> wonders if it's less work than Yahoo has been!

> Is there a current best practices how to on running one's own e-mail
> server (IMAP, not POP) out there these days? [Googling it will just
> bury you.]

One thing to remember about running an email server is that there are
two main parts to serving email, and they don't necessarily have to be
done on the same machine: you need to send mail, and you need to receive
mail.  For example, if you want to handle the email reception yourself,
you could still use a different service to send mail (such as Sendgrid
or GMail), if you don't want to worry about your server being
blacklisted.

There's also a multitude of ways that you can interface with external
servers.  For example, you could use GMail as an intermediate step to
take advantage of its spam filtering.

Anyways, for the sending end, you need to worry about
- user authentication
- making sure your server isn't an open relay (most servers these days
  should default to behaving properly)
- making sure your server is not blacklisted (there's a nagios plugin
  that can check your IP address against many blacklists)
- implementing SPF, DKIM, DMARC (optional, but helps to get your emails
  accepted)

For the receiving end, you need to worry about
- MX DNS records
- blacklists (zen.spamhaus.org is probably the most widely used one)
- whitelists (essential if you're using certain blacklists)
- spam filtering (e.g. spamassassin, crm114, dspam)
  - if you use a Bayesian filtering system, you'll need to make sure
    that you have a way of training and marking false
    positives/negatives
- greylisting (some greylisting systems integrate with spam filtering
  programs and will only greylist messages that look spammy)