[issue21935] Implement AUTH command in smtpd.

[R. David Murray]

R. David Murray added the comment:

I used some of this code in writing tests for the auth_login failure in issue 
25446 (thanks, Milan, you saved me a bunch of work :)

Now that we have asyncio and asynchat is deprecated, we've decided that the 
only real purpose of smtpd going forward is the smtplib tests.  Any other smtpd 
use should be converted to aiosmtpd instead.  Perhaps that library will 
eventually make its way into the stdlib and test_smtplib can use it as well.

So, I'm going to close this issue, but we may want to consider if and how auth 
fits into aiosmtpd (maybe as part of a 'testing server').

--
resolution:  -> out of date
stage:  -> resolved
status: open -> closed

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Roundup Robot]

Roundup Robot added the comment:

New changeset d13263ecf0c6 by R David Murray in branch '3.5':
#25446: Fix regression in smtplib's AUTH LOGIN support.
https://hg.python.org/cpython/rev/d13263ecf0c6

--
nosy: +python-dev

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Barry A. Warsaw]

Barry A. Warsaw added the comment:

Martin says: I cannot see any particular circumstances where unencrypted 
passwords for smtpd would be acceptable, given that there are perfectly 
established technologies. So I remain -1 on this patch.

Here's a use case: a testing SMTP server, such as lazr.smtptest which is built 
on top of smtpd.  In some testing scenarios, you really don't care about 
STARTTLS complications, and everything's connecting over localhost anyway.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Milan Oberkirch added the comment:

After trying to implement SMTPS with asyncore and wrap_socket I agree with 
David that it is at least hard: somehow the handshake fails 
(ssl.SSLWantReadError) and I did not really figure out why. Looking at the 
debugging output of openssl indicates that the connection drops immediately 
after setting up the session on the client side.

Anyway: I think we should apply a better version of my patch (will submit one 
soon) to be able to test smtplib (and also fix issue 8503). I'm going to make 
it clear that the AUTH functionality should only be used for testing or in 
combination with an encrypted tunnel.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Changes by Milan Oberkirch milan...@oberkirch.org:


Added file: http://bugs.python.org/file35991/smtpd_AUTH_full2.patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Martin v . Löwis]

Martin v. Löwis added the comment:

RFC 4954 states

 Note: A server implementation MUST implement a configuration in which
   it does NOT permit any plaintext password mechanisms, unless either
   the STARTTLS [SMTP-TLS] command has been negotiated or some other
   mechanism that protects the session from password snooping has been
   provided.  Server sites SHOULD NOT use any configuration which
   permits a plaintext password mechanism without such a protection
   mechanism against password snooping.

So I'm -1 on this patch, and also on the feature until STARTTLS is implemented 
(and then this patch needs to be updated to conform to this requirement).

--
nosy: +loewis

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Milan Oberkirch added the comment:

My interpretation of this paragraph is the following (English is not my native 
language so please correct me if I'm wrong):
The requirement is to provide a configuration where plain auth is disabled if 
password snooping would be possible otherwise not to forbid such configurations 
generally. An admin SHOULD provide security measures to prevent password 
snooping. 

Setting enable_AUTH=False is a configuration where plain authentication is not 
permitted. The admin should provide a STARTTLS (or any other encrypted) tunnel 
if enabling AUTH (stunnel would be a common solution on Linux).

Maybe we should explicitly mention that in the docs?

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[R. David Murray]

R. David Murray added the comment:

Providing starttls support would be the preferred solution, but that is a Hard 
Problem.  We probably need to rewrite smtpd using asyncio in order to provide 
starttls.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Martin v . Löwis]

Martin v. Löwis added the comment:

Milan: Your interpretation of the MUST requirement is correct.

However, we still cannot support the SHOULD NOT requirement: A server operator 
SHOULD NOT accept unencrypted passwords. RFC 2119 explains

   This phrase, or the phrase NOT RECOMMENDED mean that
   there may exist valid reasons in particular circumstances when the
   particular behavior is acceptable or even useful, but the full
   implications should be understood and the case carefully weighed
   before implementing any behavior described with this label.

I cannot see any particular circumstances where unencrypted passwords for smtpd 
would be acceptable, given that there are perfectly established technologies. 
So I remain -1 on this patch.

A (not recommended) STARTTLS alternative is SMTPS (port 465). I would be -0 if 
there was an SMTPS implementation in smtpd, and the documentation would discuss 
that AUTH is best used with SMTPS until STARTTLS is implemented.

I don't understand why STARTTLS would require asyncio. Wouldn't wrap_socket 
solve the problem?

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[R. David Murray]

R. David Murray added the comment:

I haven't looked at the problem myself.  Someone (Giampaolo?) told me that 
wrap_socket wouldn't work because of the fact that smtpd uses asynchat.

As for the 'particular circumstances' clause, I would suggest that one of the 
primary use cases for smtpd is in testing infrastructure (and that is in fact 
the motivation for this issue, for stdlib testing), and in that situation 
unencrypted passwords would not be an issue.

However, if wrap_socket/SMTPDS would work, it would make the feature useful 
beyond the testing arena.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Milan Oberkirch added the comment:

Done. I added the keyarg 'enable_AUTH' and two abstract methods to the server:
process_auth(user, password) for authentication and accept_recipient(user, 
mailfrom, rcptto) for authorization.

--
Added file: http://bugs.python.org/file35974/smtpd_AUTH_full.patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Milan Oberkirch added the comment:

We could solve issue 8503 at the same time by always calling 
'accept_recipient(mailfrom, rcptto, user=None)' and providing a default 
implementation for backwards compatibility.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

Milan Oberkirch added the comment:

There is no real API in the current patch and authenticating has no effect 
(other then preventing you from authenticating again and storing the username). 
I am wondering how the user should turn AUTH on/off.

Solution 1:
add a keyword argument 'enable_AUTH' and require the programmer to override 
_verify_user_credentials. This function could
1.1 raise NotImplementedError
1.2 deny access
by default.

Solution 2:
add a keyword argument 'authentication_function' which turns AUTH support on 
when given and provides the function used to verify user credentials.

Solution 3:
enable AUTH if self has the _verify_user_credentials-function as attribute (and 
leave it undefined in the base class)

I think solution 1 is the most explicit so I'll implement that so we have 
something to discuss :)

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[R. David Murray]

R. David Murray added the comment:

Describing how a programmer would implement authentication is exactly the API I 
was referring to, and that includes the signature and semantics of 
_verify_user_credentials.

I agree that (1) seems the cleanest.  I'd favor 1.1, NotImplemented, which lets 
the programmer see immediately what they did wrong.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[R. David Murray]

R. David Murray added the comment:

I think it would be a good idea to write the documentation.  It is much easier 
to get a feel for the API via docs than it is via code.  (That is, when you 
explain how to use the API, you sometimes find design bugs :)

for a/b: so you are thinking of an auth function passed in as opposed to a 
method overridden in the subclass?  That is consistent with how we enable other 
capabilities, so it has some attraction.

I haven't gone over the code in detail yet, but it looks like you need to add a 
reset of 'self.user' during the applicable state transitions.

--

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue21935] Implement AUTH command in smtpd.

[Milan Oberkirch]

New submission from Milan Oberkirch:

I implemented message processing for LOGIN and PLAIN authentication in smtpd. I 
also patched test_smtplib to make use of this functionality.

The goal for the API is to provide decryption and message processing in the 
smtpd library and call a externally provided or overwritten function(user, 
password) to verify credentials.

The patch provided is missing documentation and a proper API to use/activate 
this feature (and API specific tests of cause). Things which need to be 
discussed:
a) how shoud AUTH be activated? (when should MAIL etc. require AUTH? when 
should 'AUTH' be announced in the EHLO response?)
b) how should the programmer change the _verify_user_credentials method?

My idea to solve a and b at once would be to set the verification function by a 
keyword argument and require/activate AUTH if this kwarg is set.

I didn't implement CRAM-MD5 because it requires the correct password to be 
available in plain text.

--
components: email
files: smtpd_AUTH.patch
keywords: patch
messages: 222516
nosy: barry, jesstess, pitrou, r.david.murray, zvyn
priority: normal
severity: normal
status: open
title: Implement AUTH command in smtpd.
type: enhancement
versions: Python 3.5
Added file: http://bugs.python.org/file35895/smtpd_AUTH.patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue21935
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com