From: (no name) (no email)
Date: Sun Feb 01 2004 - 10:56:37 EST
On Sun, 1 Feb 2004, Wietse Venema wrote:
> 1) a) Specify "mailbox_command = /path/to/maildrop..." in main.cf.
> b) Specify "mailbox_transport = maildrop" in main.cf
> c) Specify "fallback_transport = maildrop" in main.cf
>
> 2) a) Specify a transport map
> b) Specify "local_transport = maildrop"
>
> MAILDROP_README originally discussed variant (2) only. Tonni's
> comments, on the other hand, are applicable to variant (1).
>
> The old LMTP_README document mixed up variants (1) and (2). having
> fixed that, I am now faced with an LMTP_README that is completely
> at odds with VIRTUAL_README, and with a MAILDROP_README that gives
> an incomplete account of the same things that LMTP_README discusses.
> This still needs to be reconciled.
>
IMHO, the real issue is that alternative delivery agent integration
deserves its own README (LDA_README?) which can be referenced from
multiple places. The LMTP_README, MAILDROP_README, ... should largely
remain about using LMTP or MAILDROP as full transports. References should
be given to the LDA_README (if that is a good name) document for how to
integrate an arbitrary delivery command with the local delivery agent.
LMTP is somewhat special in this regard, because it is a protocol not a
command, so there is not necessarily a "mailbox_command" for LMTP,
although it may be possible to use the Cyrus "deliver" as a command-line
LMTP injector (never tried it) to non Cyrus LMTP servers.
If one really wants to talk about the database support for say MAILDROP,
the issue is that in the virtual mailbox domain case, the "maildrop"
transport does not appear to support the Postfix "@domain" catchall
mechanism. This is done by Courier-MTA instead. So if one wants to use
maildrop as a VDA for a domain with a catchall, one needs to use
virtual(5) with a catchall entry and identity entries for all real users.
This perhaps should be added to MAILDROP_README, along with a reference to
VIRTUAL_README describing the setup of virtual mailbox domains in general.
For me the sample-local.cf was sufficient documentation in the absense of
LDA_README, but it is likely a good idea to provide a more chatty version
of the sample-local.cf content, with 2-way references to VIRTUAL_README,
LMTP_README, MAILDROP_README, ...
Indeed LMTP_README is not the most obvious place to search for information
on alternative final delivery mechanisms (when the mechanism in question
is a delivery command rather than the LMTP protocol).
-- Viktor.
|
|
|