Hi,
I understand that we need to fix im-switch and simplify it. In order to
do it, we need to have good design first. .uim actually created by the
script in uim-anthy(source uim) package.
My current thought is that :
im-switch switches between
uim *
scim *
kinput2 - canna
kinput2 - wnn
ami
cin
...
(Trying to fiddle uim or scim setting from im-switch is bad idea)
Anytime user want to configure uim or scim, he/she can use gui tool.
im-switch should not interfare with it. im-switch can be instrumental
by giving suggestion message on backend conversion tools but it should
not toutch which backend to be activated.
As for symlinks, it was added to enable any locale to be able to
activate uim or scim as default IM if sysadmin wants to do so. Maybe
this was not elegant solution of mine but it works. See below for
rationale.
On Mon, May 15, 2006 at 10:53:59PM +0200, Jan Willem Stumpel wrote:
> Osamu Aoki wrote:
>
> > /usr/share/doc/uim-common which is symlinked from
> > /usr/share/doc/uim-gtk2.0 too.
>
> Yes, I found it now. So there are two preferences methods, the
> graphic one which uses .uim.d, and the other one which uses .uim.
>
> But the graphic one is installed by default, so we do not need the
> other one. It just overrides the user's own preferences.
Well this is true if you indstall uim . Just installing uim-anthy does
not install GTK tool bar.
> Example:
>
> -- A user installs uim (which automatically installs im-switch),
> anthy, and uim-anthy.
You mean "install uim and uim-anthy". I guess optimization of
dependancy around uim and scim package is good idea.
> -- The user is clever enough to know that he/she has to call
> im-switch -s uim_anthy before starting X, and to run the uim
> toolbar (because im-switch does not take care of that).
> (Really there should be a dialog offered when you install
> im-switch).
This should not be needed if X is running under ja_JP.UTF8 and other
typical locales where IM is normal.
For other locals (en_US, ...), you need to add it as user using
im-switch -c
or
im-switch -s uim-common
(since I am proposing dropping uim-anthy, uim-common provide this.)
If system wide default needs to use uim or scim, maybe there should be a
way to do it.
> -- The user decides he/she wants to use another default input
> method in uim, not anthy (e.g. 'direct'), using the uim
> Preferences menu. It won't work. At the next X restart, anthy
> is still the default method (because of the .uim file). The
> 'direct' method, (and the 'byeoru' uim method), can
> not even be got through im-switch -c or im-switch -s, so you
> cannot make them the default at all.
Yes I know. As long as we do not touch .uim, we are fine :-)
> Let us not make life too difficult for users. Let us get rid of
> this .uim file, please. It is not needed and it only causes trouble.
I know. I initially thought the same but Muto-san had different
opinion. It was him. My recent upload was bug fix and code clean up
only. I did not change fundamental design of Muto-san.
> > But please understand the actual "buggy" codes per your
> > assessment are not in this package. They are example.
>
> I don't understand. Surely creating the .uim file is in the package?
/etc/X11/xinit/xinput.d/uim-anthy is installed by uim-anthy package
[EMAIL PROTECTED]:/etc/X11/xinit/xinput.d$ dpkg -S uim_anthy
uim-anthy: /etc/X11/xinit/xinput.d/uim_anthy
This file was decided to be installed based on the im-switch
documentation. This file (program) is called from im-switch installed X
start up code to create .uim.
> > [..]You do not need to be root to set up like you did if you
> > are fixing for one user.
>
> I did that only to cure the bug (= the fact that im-switch keeps
> re-creating the .uim file).
Well, I mean You should not ... if I fix this issue from uim-anthy.
> > im-switch has 2 modes, user and root. In user mode, you copy
> > uim-anthy to ~/.xinput.d and edit.
>
> You mean ordinary users have to edit a config file by hand in
> order to use im-switch?
>
> > Run "im-switch -s uim-anthy" to point to user one. New -l
> > comand should tell you did OK. See my new documentation in
> > README.Debian(unstable).
>
> I just got the new version now. It is even more complicated! There
> are now symlinks for 170 legacy locales in
> /etc/X11/xinit/xinput.d. And they point to /etc/alternatives,
> where there are again symlinks. Half of /etc/alternatives is
> now used up by im-switch alone! Are you sure this is necessary?
> Then, almost all of these symlinks point back at one file:
> /etc/X11/xinit/xinput.d/default.
Yep. That is why all local has choice. Even if SCIM or UIM is
installed, we can keep en_US not defaulting to uim/scim but make ja_JP
and zh_CN defaulting to uim/scim. This is the reason behind many
symlinks.
> Why would im-switch have to pay any attention to the user's locale
> at all? If people have a legacy locale, they select a keyboard
> (when they install Debian) for that locale.
If you start X in en_US.UTF-8, you expect no funny IM thing.
Many European people may be using simple X based deadkey which requires
no SCIM/UIM as default.
If you start X in ja_JP.UTF-8, you expect some IM thing. (SCIM/UIM)
Although this sety up should work with legacy locale such as
ja_JP.eucJP, we care locale only for above reason.
> Or perhaps they
> install a CJK system like kinput2 or anthy (without uim). No
> multilingual input in that case.
kinput2 works with UTF-8 if I dare to do.
> But if people want to use really multilingual input, it does not
> matter what their locale is (but it has to be UTF-8, otherwise uim
> won't do anything).
Yes. But we do not know if they want multilingual input. That is why we
need to guess their preference based on locale and that preference
should be configurable in system wide level and user wide level.
> IMHO im-switch should help to install multilingual systems like
> uim, scim, and IIIMF.
Yes. And I think it does it.
> It should do that in any (UTF-8) locale the
> user happens to be in.
Actually, just because you are in UTF-8 environment, we never know what
he wants. This should be configurable.
> Then the user can use uim's, (or SCIM's, or
> IIIMF's) preferences menu to select which input methods he/she
> wants, and which should be the default one.
I can not agree more. Muto-san, let's drop support on configuring how
UIM is congigured to use its backend. That can be done just updating
documentation and filing bug to uim and uim-anthy package.
> So im-switch can really be drastically simplified. I'll help you
> do this if you want. All this idea of 'locales' should be thrown
> out. Then the doc can be simplified, simply saying
>
> -- if you want uim, do this
> -- if you want scim, do this (etc.)
Yes sure. But let's keep other options such as ...
-- if you want kinput2 in en_US.UTF8, why not :-)
> Then the uim/scim should come up at the next X restart (which it
> does not now in the case of uim).
If we fix uim-common package, it will do so for ja_JP, zh_CN,...
For en_US, if you configure uim being system default, it will.
im-switch package needs no modification except developer guide part of
documentation.
> Ideally, im-switch should pay
> attention to the window manager which is in use (icewm for
> instance needs some special treatment; maybe others too).
What are these? uim start up mode. systray application?
Osamu
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
