Depends on how you think about it.
If you forget about email addresses. Dovecot works on mailbox's, and
it maps a mailbox to a path, username, and password.
In postfix, it only cares about email addresses (if you use dovecot
for delivery, if not then postfix also has to care about the mailbox
location). In this case you just tell postfix the email addresses that
are valid, and what mailbox they go to.
So normally most users would just have a 1 to 1 mapping in postfix,
email -> email, as their email address will be the same as their
mailbox. But then you might have extra, like, sales@x -> user@x
All depends on how flexable or simple you want it later. You could
just manage two flatfiles. Or you could have it create the 1 to 1
mapping automatically with a script, and just do the extra mappings
seperate. Or do the whole thing in sql, and use like postfixadmin to
manage it all. Or even use postfixadmin, and have a script pull the
results into flatfiles that it uses.
It all depends on how much time and energy you want to spend in
setting it up, vs the flexibility you in vision you need later.
I do it 3 different ways, on different systems, one is just sql fully,
nothing interesting. My personal email is sql, but dumped to local
flatfiles. And another system I pull the info from windows AD.
Quoting terryjames9461@mm.st:
Hello Patrick,
On Tuesday, September 27, 2011 6:06 PM, "Patrick Domack" patrickdk@patrickdk.com wrote:
Using dovecot lda/lmtp you remove all postfix needs to know mailbox name to directory mapping, that would be duplicated.
With using the Dovecot lmtp option, where does Postfix know to refuse email for a non-existing user or domain? That also has to be shared?
I am trying to draw a picture in my head of all the data pieces. Are you saying that when using lmtp the data for Postfix and the data for Dovecot/LMTP do not overlap anymore? Each can have its own flatfiles?
TJ