[Uim] Design philosophy and strength of uim

[Ken Deeter]

Just my 2 cents:

I think it may be worthwhile to try to see what overall goals that need
to be achieved in terms of input methods and see how each project fits.

In my mind there are several clear and related goals:

1) Create a system that works now, is multi-lingual, has some kind of
switching mechanism, and is available ASAP.

2) Create a comprehensive 'framework' for input methods, including
thinking about how to manage them, how to provide consistent GUI's, how
to unify access to them (both from the user's point of view and also
API point of view), and how to keep their development manageable.

3) Explore new ideas in input methods, including new API's and new
interaction mechanisms.

All the efforts right now contribute to each of these goals in different
ways.

As for goal #1, it looks like combination of SCIM + IIIMF + uim will
work the best. To meet this goal, each party needs to make sure that
they ahve modular and stable designs so that the functionality of each
system can easily be integrated with the other. The scim-uim efforts as
well as the scim-iiimf efforts fall into this category, I would say.


As for goal #2, it seems that SCIM and IIIMF have different approaches,
but the idea is similar. The strengths of these frameworks are that they
have been carefuly built with this type of goal in mind, and should
continue to be developed with that goal.

For Goal #3, I feel that uim and maybe to some extent scim fit well. For
uim in particular, I think that what this means is that uim developers
can worry less about how to implement IM switching or desktop IM
management (or legacy compatibility with XIM) and rather explore new
avenues for input method API's. Even if it means just developing novel
methods for Japanese, in the end, any new API concepts developed in UIM
should cause IIIMF and SCIM to react if their API's are unable to deal
with the new requirements.

In short, we need to separate the goal of 'integration' with the goal of
'exploration' and realize that the three projects can reduce a lot of
overlap if it is made clear which direction each is heading in.

As a recent new user to scim, i am extremely greatful for the
integrative capabilities it has, and cannot wait for the qt-immodule
version. I would very much like for scim to continue its efforts on
trying to figure out how to provide uniform interfaces to input methods,
and to maintain bridges to new methods like uim. At the same time I
would like to see uim keep forging ahead trying new ideas, without the
fear or risk of destroying SCIM's structural stability just to add some
small little feature that actually might not be too important in the
end.

I would say likewise for IIIMF, though I have not been able to use it
directly on my gentoo box yet (even though I wrote the original versino
of the pacage ;-P).

Ken