Re: 3.6 Feature: IBus/XKB integration

[Owen Taylor <otaylor redhat com>]

On Wed, 2012-04-25 at 16:00 +0100, Bastien Nocera wrote:
> > Isn't that a whole lot more confusing? Plus, do we have time in the 3.6
> > cycle to do a good job at inventing a replacement for non-legacy apps?
> > Is that the best use of our resources?
> 
> That's an example. And from what I understood, given the "need to
> mention the iBus chosen method through envvars" system, I think we have
> part of the answer here. We're already effectively working on the
> non-legacy replacement.

 If I'm entering *text* into an text field, I need an input method

 If I'm pressing a hotkey, an accelerator, a mnemonic, etc, then input
  methods don't matter, and the toolkit needs to know something about
  the keyboard layout.

(For example: there's no way that the code in gnome-settings-daemon can
ever use an input method to figure out which keycode it needs to grab on
the root window to implement a hotkey. And furthermore, if the user sets
one of these hotkeys to Control-Alt-G, that hotkey can't stop working
because the user switches to another layout without a G in it.)

Just because we're using input methods now, doesn't mean that keyboard
layout knowledge is legacy. So, if you declare XKB legacy, you need to
invent something else.

If somehow more than 4 keyboard layouts has become an essential feature,
then I suspect the approach of filling the 4 XKB slots with:

 - The current layout
 - The layout for the UI language
 - The Latin layout
 - And one other layout

Probably works close to OK.

- Owen