[fdo] Standardization of input methods on Linux?
David Oftedal
david at start.no
Wed Sep 13 17:13:14 PDT 2006
Hello!
I don't know how well this fits within the scope of Freedesktop's
current work, but as a Linux user who studies languages, I think there's
an area that could definitely use improvement, namely input methods for
various Asian and other non-romance languages. While several solutions
for inputting these languages are available, most of them seem to be of
the "rocket science" variety rather than the "desktop OS" variety.
Currently, there are several incompatible interfaces for implementing
input methods on Linux. The most universal one may be XIM, or X Input
Method, which unfortunately requires some unnecessary configuration and
doesn't seem to provide any facility for switching input methods within
a program once it's been started. GTK+ 2 and Qt 4 both appear to provide
their own seemingly incompatible standards, whereby an input method can
be made to work only in GTK+ 2 or Qt 4 with minimal configuration.
Finally, there are third-party solutions like SCIM, UIM and IIIMF, which
aim to unify the pre-existing interfaces into one, handing input methods
via software that in turn uses one of the pre-existing solutions to
communicate with programs. These unfortunately don't always work
satisfactorily and can be hard to set up for some users.
Some input methods also seem to have problems coexisting with keyboard
layouts (for instance, the non-romanized Korean keyboard layouts may
assume a QWERTY keyboard and thus not function with other layouts), only
a select few seem to work on the console (I remember a "canuum" or
something similar for Japanese), and the various methods of switching
between, activating and deactivating input methods are currently so
diverse that it frequently gets confusing.
It should theoretically be possible to create a more unified solution
that works at least across X and across all keyboard layouts with zero
configuration, more similar to what one can see in Mac OS X and Windows.
However, it would require some sort of standard to be worked out with
the aid of a lot of very competent programmers. Perhaps this is the sort
of thing Freedesktop could do after all?
Best regards
David J. Oftedal
More information about the freedesktop
mailing list