180806 - IMAP backend should avoid unnecessary logout

Status: RESOLVED FIXED

(MailNews Core :: Networking: IMAP, defect)

Description

[Kai Engert [:KaiE:]]

I'm not perfectly sure what I propose is a good idea.

Suppose you are connecting to a server using IMAP/SSL.
Suppose your configuration is to use client authentication with the server,
where the authentication happens on the SSL layer around the connection.
Suppose when the user opens the connection, the prompt comes up that asks the
user to enter the password to unlock self's certificate for authentication.
Suppose the user decides to cancel.

In the above scenario, when the user quits the application, one can see the mail
code is attempting another connection to the server on exit.

This happens because the imap protocol code is trying to send a
 "1 logout"
command, which agains triggers the attempt to establish a crypto communication.

I want to suggest we prevent sending the logout command on closing, if it would
be the first command to the server.

Comment 1

[Kai Engert [:KaiE:]]

Attachment: Proposed Patch

Small change, large context, patch ignoring whitespace.


What do you think?

Comment 2

[David :Bienvenu]

Comment on attachment 106719 [details] [diff] [review]
Proposed Patch

can we get authenticated without ever sending a command? e.g., the pre-auth
stuff? Even if we could, I guess we'd immediately try to send some commands
anyway, so it might be a moot question. If that seems OK to you, then it's fine
with me.

Comment 3

[Kai Engert [:KaiE:]]

> can we get authenticated without ever sending a command? 
> e.g., the pre-auth stuff? 

Yes. The SSL authentication completes without having sent any IMAP command. An
IMAP server can query its crypto engine, whether authentication already happened
for the given server - and that is why one does not see a password prompt when
connecting to our corporate internal mail servers with a personal certificate
installed (or when going to aka using https).


> Even if we could, I guess we'd immediately try to send some 
> commands anyway, so it might be a moot question. 
> If that seems OK to you, then it's fine with me

You sound like you are concerned about not sending a logout command although we
are authenticated with the server using a non-IMAP protocol mechanism?

I suppose, the worst that can happen is the server continues to have us
authenticated as long as we keep our socket up.

Comment 4

[David :Bienvenu]

Kai, yeah, that's what I was worried about, but I think we'll pretty much always
send at least one command to an imap connection - in other words, we only open
connections in order to send commands to them.

Comment 5

[(not reading, please use [e-mail] instead)]

Comment on attachment 106719 [details] [diff] [review]
Proposed Patch

sr=sspitzer

Comment 6

[Kai Engert [:KaiE:]]

Fixed on trunk.