Re: On the spot input

[Owen Taylor <otaylor redhat com>]

NIIBE Yutaka <gniibe@chroot.org> writes:

> With regards to XIM, I think that it would be better _not_ to depend
> on the implementation of X.  I think that it's kind of kludge.
> 
> I believe that input method support should be designed as one of
> editing facilities (such as deleting character, killing word, and mark
> the part of text...), not as some kind of features of keyboard input.
> 
> How about defining some editing extentional API for GTK entry and GTK
> text?  And implementing input method support with the extention?
> Possibly, the extention uses CORBA for communicating conversion
> service or dictionary support.
> 
> I proposed this solution last year, but I've not worked at all.
> So, no working code yet.
> 
> The idea is extention to readline library for TTY case, and extention
> to GTK for window environment.
> 
> Any comments are appriciated.

I'd agree that XIM is a kludge, and not very nice. However, supporting
it is definitely a requirement. We need to be able to work with 
existing input methods.

It could be interesting to support loadable input-method modules in 
GTK+.

I'm thinking of making the GdkIC (input context) structure a GTK+
object, so we can use signals for input callbacks. If we made the
GtkInputContext object just a "base class", then different modules
could load up different implementations of the base class:

 - XIM based
 - CORBA based
 - etc.

That's my current thoughts on the issue.

Regrads,
                                        Owen