[Dovecot] Why deliver+usercheck? deliver+MTA?

[Lukas Haase]

Am 13.10.2010 13:08, schrieb Daniel Luttermann:
> Lukas Haase wrote on 10/13/2010:
> [...]
> By default, Postfix rejects mails for unknown local users.If Postfix
> accepts mails for unknown users than it's a configuration problem or
> you don't maintain a list of valid users.

Yes, but I am talking about virtual users.

>> Is there a special reason why there is no discussion about this?
>
> It's Postfix related - Dovecot does no checks about valid recipients
> for Postfix but you can use the same data sources as for Dovecot - no
> need to maintain user lists for Postfix and Dovecot.

But *why* would you want to let dovecot (deliver) check this?

In any reason the MTA *must* have validated the existance of the local 
part. I do not know any reason why deliver should do this.

And again: Both
http://wiki.dovecot.org/LDA/Postfix
http://wiki.dovecot.org/LDA/Exim

describe setups for virtual users. But none of these pages give a hint 
that the MTA needs to check the local part too.

> Because Postfix needs to check for valid recipients why should there a
> special hint in the Dovecot Wiki about that?

Because if someone implements a system based on the WIKI above he builds 
up an insecure system (producing backscatter).

> You must first make sure
> that Postfix works as expected - no other IMAP Server checks vor valid
> recipients.

Yes but no other IMAP server (but I only know Courier!) checks the 
validity of the user in the LDA. maildrop for example does not.

>> However, as postfix seems to be really too unflexible I have set up exim
>> to handle incoming mail and do the usercheck in the router (with an LDAP
>> query). But now the user is doubled-checked: Once when receiving with
>> exim and a second time in deliver. This is not necessary, so I guess I
>> can disable the LDAP query for deliver and set up a static userdb.
>
> Why is Postfix unflexible? Use reject_unverified_recipient for dynamic
> verification of valid recipients and there's no need to maintain
> static files. You could also use a LDAP query to retreive a list of
> valid recipients before you accept the mail for non-existing users.

Thank you! Does reject_unverified_recipient also work when the mail is 
passed to deliver as described in http://wiki.dovecot.org/LDA/Postfix 
"Virtual Users"? If this would be the case then this is exactly what I 
was looking for!

Until now I tried to use an LDAP query. But also deliver uses an LDAP 
query to check the existance of the user. And this was my question if 
both of them are necessary.

To the question why postfix is too unflexible: I found no way how to:

* Hook up *fully* virtual users with dovecot (using deliver) for domain 
example.com
* Hook up mailing lists for domain example.com using mailman

The current setup uses system users and therefore this setup is no 
problem. But now there are virtual users ...

>> Why does the Wiki recommened to verfify with deliver when the user needs
>> to be checked at the MTA anyway?
>
> Checking of valid recipients is a Postfix job so you can use
> relay_recipient_maps, reject_unverified_sender or virtual_mailbox_maps
> (depending on your configuration).
>
> Btw: what does the Wiki recommend? Weblink?

Yes of course, it is a postfix job. But also postfix jobs are described 
in the Wiki: http://wiki.dovecot.org/LDA/Postfix. And I think a small 
hint that the user must make sure that local parts are validated would 
be fine.

Regards Luke