On Mon, Dec 13 2010, Lars Magne Ingebrigtsen wrote:

>> I don't think assistants is a good solution. That would just hide a
>> complexity. But things should not be complex.
>
> They shouldn't, but sometimes they are...  That is, it's more pleasant
> for the user if we query the user instead of making a wrong guess, in
> general. 

Just don't forget who your users are. :)

> After the `g' rework, the only difference real is in the group name.
> That is, the primary method groups don't have prefixed names, while the
> rest do.  We could hide that by not displaying the prefixes (by default)
> in the group buffer, but it's a leaky thing...  and doing a total rework
> here is ... eh...  Kinda a lot of work.

I agree it's a lot of work, but it's not undoable. At least, we should
decide that it would be a good idea to revoke the primary/secondary
mechanism and find something else.
Then we can do the work little by little. I volunteer. :)

> instead of exposing the primary/secondary stuff directly.  I mean, if we
> just drop the idea of a primary method altogether, and just say that
> they're all secondary, then everything is fine.  Except for the long,
> prefixed group names...

Sounds like a plan!
Or we could provide a prefix for each method, just without the backend
name in front of it. So you can still set it to "" if you have one
primary, and to "WhatEver" so you can prefix the other methods.

>> Weird, I do not have one. It may be recent.
>
> Perhaps you've killed it?  It's only relevant for Agent users.

Ah, I do not use the agent since it has a tendency to not work with
IMAP.

> However, that's really not the proper fix, since native groups (again
> with the difference) aren't prefixed.  I'll add a different fix.

I don't know what fix you want, but setting it to "." might be the best
fix. :)

-- 
Julien Danjou
❱ http://julien.danjou.info