[Mailman-Users] virtual domains

[Stefan Förster]

* Mark Sapiro <mark at msapiro.net> wrote:
> Melinda Gilmore wrote:
> 
>> I know this has been covered alot and I do see alot of talk in the
>> archives, but I am new to the whole, mailman/postfix/apache/redhat
>> world and the instructions are not clear to me because of that.  Is
>> there anyone out there that has the actually settings and what files
>> to change in postfix, apache, mta, and whatever to get virtual
>> domains to work.  Am I to understand that if I had a list called
>> something at lists.service.osu.edu (our domain) and I want a virtual
>> domain to go with it called something at network.org I can get this to work?
> 
> 
> Not exactly. If you have a list named "something" that is currently in
> the lists.service.osu.edu domain, you could move it to the network.org
> domain, but the whole idea of virtual domains in Mailman is that the
> various domains are disjoint (although currently list names still have
> to be globally unique).
> 
> If you just want to be able to mail to something at network.org in
> addition to mailing to something at lists.service.osu.edu and have either
> address go to the list, this would all be done in Postfix and DNS.

I think one would have to add something at network.org to
acceptable_aliases, too.

As a remark, I really think that integration of Mailman into Postfix
using relay_domains and postfix-to-mailman.py should be included in
the official installation documentation. Postfix's concept of
virtual_alias_domains is very poweful and therefore offers lot's of
ways you can totally break your mail system if you are inexperienced -
you can break recipient validation or introduce mail loops, for
example.

Any aliasing needed can still be done with virtual_alias_maps if one
absolutely _has_ to have aliases. With relay_recipient_maps you get
better control about acceptable recipients (although you have to set
this up yourself, I didn't see any way in Mailman to handle this in an
automatic way).

And as a last advantage, on a busy server, I personally would refrain
from using Berkeley BD hash tables, because updates to those are not
done "atomically" which might seriously break mail to lists if mail
arrives while the maps are updated. Using CBD tables help's a lot
here, safe for the much smaller memory footprint.



Cheers
Stefan
-- 
Stefan Förster     http://www.incertum.net/     Public Key: 0xBBE2A9E9
Science is what happens when preconception meets verification.