Sorry, Noel,
Now that I re-read your last post, I can see there is no discrepancy at
all between my findings and your description in the two cases I mentioned.
In fact, what happens is exactly what you describe. The email message is
rejected because the client specifies a MAIL FROM listed in
smtpd_sender_login_maps and the client is not logged in.
I mis-interpreted your explanation and thought you meant that the client
was rejected after examination of ownership in smtpd_sender_login_maps,
while in fact ownership is not checked. Your explanation was accurate.
Sorry for the complication, and thanks again for your detailed analysis.
I believe this explanation should find its way into the documentation,
because I am afraid the explanation of the three directives is otherwise
obscure.
All the best,
Nick
On 11/2/2011 5:58 μμ, Nikolaos Milas wrote:
Thanks Noel, for the detailed info.
In the meantime, I had already tested, and here are the test results,
for reference (tested by removing ownership of f...@example.com by foo
and logging in (in scenario II) as user foo):
I. 1 --->a (message rejected; user asked to login), 2--->b (message
accepted; ownership not checked), 3--->a (as in 1)
The error in cases 1 and 3 is:
553 5.7.1 <f...@example.com>: Sender address rejected: not
logged in
II. 1 --->b (message accepted; ownership not checked), 2--->a (message
rejected due to ownership), 3--->a (as in 2)
The error in cases 2 and 3 is:
553 5.7.1 <f...@example.com>: Sender address rejected: not
owned by user tester
So, in essence, if the client is unauthenticated, I've found that the
smtpd_sender_login_maps will never be checked directly.
The above test results differ from what you thoroughly explained in
cases I.1 and I.3, where ownership is not actually checked, yet the
message is rejected because the client is unauthenticated.
Could some other directive affect the observed differentiation?
Nick
