What to read about interactions

[Steffen Kaiser]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, 5 Apr 2015, Andrew Sullivan wrote:

> I've been using dovecot happily for some years, but in a recent
> migration of my mail server I set up something that I think will be
> slightly easier to maintain.  I now have a single (in my case
> postgres) database backing all user names and so on, with postfix
> relying on dovecot auth for user validation and delivery.  I have the
> usual anti-spam machinery in there, and have started using the (for me
> new) pigeonhole implementation of sieve.  (My old mail system was
> built around the earlier dovecot implementation, and while I've
> patched it over time it's really pretty creaky.)
>
> On the whole, this all seems very nice, but I'm having a hard time
> understanding the details of what interacts with what other bits when.
> In particular, my mail system is known by several different names, and
> I thought it'd be much handier to have everything land in one real
> mailbox, since I could sort the inbound mail into different mailboxes
> without a lot of trouble using seive.  This has turned out to be
> slightly trickier than I thought it ought to be (procmail is in some
> ways easier), and I've come to the conclusion that I don't have a
> strong foundation in exactly how this is supposed to work.

procmail is easier ;-)

Think of "Sieve" taking over procmail.

If you want to know how to write Sieve scripts, check out 
http://wiki2.dovecot.org/Pigeonhole/Sieve/Examples or any other Sieve 
site. If you want to filter by mail domain, you need envelope tests, see 
http://wiki2.dovecot.org/Pigeonhole/Sieve, too.

e.g.

if envelope :is "to" "owner-cipe-l at inka.de" {
   fileinto "lists.cipe";
   stop;
}

> Apart from reading the code (which would probably at this point be
> hard enough for me so as to make it impossible), are there things I
> ought to read to understand this better?  I've read the seive RFCs,
> but while that gives me a clue about what ought to happen
> protocol-wise, it isn't helping me at all to understand the
> interaction among the different subsystems so that I can see clearly

There are no "subsystems", as far as I understand the term. There is 
information about the current message, which can be tested and acted on.

> how different parts of the delivery chain are interacting.  I'm sure
> there's some crucial bit of, "Everyone knows that," that I've
> overlooked, and I thought it would be better to ask for the FM than
> just to read randomly until I stumbled on it.

- -- 
Steffen Kaiser
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEVAwUBVSPQAnz1H7kL/d9rAQKwzggAgmpsAzrm0FE3/TVMQVq3tnjqxoFwX6ur
JhO1gb1558bEhv95rwtcmsPcEPP8fqstcMoZv9rKnPjgorAzx+t9/hxNhvzer49U
q2NcRRLBrWKt/GdJK3f06Ifs+tUjuog1kotiyR5E193xf/cvEd98H4EyVR35kak3
3tDHsW7ylhU7bg3N6FU9fu6px8EsVKQFG/H3mClltpgsgsq5AJ1jMM6giZH2BJ9d
IwzpnP8WmvXOpTt6YvJOLySiAjWY+8e0c/LjNt/8dmOL/rdqKGY70RNopOOnGarC
JBh0xxDaKnqv3mwbhFKEVLtr+qhKTGbDjlF1G49CebpNpIsTTIXZIA==
=0pPx
-----END PGP SIGNATURE-----