Re: nnimap-user setting?

[Ted Zlatanov]

On Mon, 18 Apr 2011 16:12:45 -0300 Stefan Monnier <address@hidden> wrote: 

SM> Doesn't sound right: what if you have several accounts on the same
SM> IMAP server?  auto-info needs to know the username in order to find the
SM> right password.
>>>> In Gnus, you'd set up multiple secondary servers with the same machine
>>>> address and each one has its own nnimap-user.
SM> That's one way to do it, but not the only way.  The user may only want
SM> one of the accounts in his list of secondary servers.
>> Can you give a specific example of what you mean and how it should work?
>> I may be missing your point.

SM> The autoinfo file has data for address@hidden and address@hidden, and the
SM> user wants to setup his .gnus to connect to address@hidden and nothing
SM> else (maybe he has another .gnus elsewhere configured to only connect
SM> to address@hidden).

I see.

>> You always put `:user nnimap-user' in the `auth-source-search' spec.
>> That means that a line like "machine x password y" won't work if
>> `nnimap-user' is set.  It's not annoying, just
>> unnecessarily restrictive.

SM> I'd expect a ":user nil" argument to work just like when there's
SM> no :user keyword argument, so as long as the user doesn't set
SM> nnimap-user the behavior should stay unchanged.

You're right.

>> Also if the `nnimap-user' is set it should override anything else coming
>> from `auth-source-search'.  I think your patch allows it to be
>> overridden.

SM> I'm not sure what you mean here.  If I specify ":host foo.bar :user
SM> titi" I would expect auto-source to ignore an entry for address@hidden
SM> So either auto-source works differently than what I expect or you're
SM> talking about another case of override.

s/auto-source/auth-source/g :)

Your patch is correct.  I was thinking incorrectly about the :user nil
search case, which will DTRT as you said.  Sorry for
the... interjection.

Ted