Re: [RFC] Text Input Management System (5)

[Kazunobu Kuriyama]

Chad Hardin wrote:

> I looked it up and realized it is an X11 extension, correct? 

I'm not sure what you mean by this.  In case you refer to XIM, I don't
think we can call it an extension of X11 because it dates back to the
era of X11R5.  Also, as a matter of fact, the code I wrote doesn't use
any X11 function directly at all.  So I don't understand how you reached
that conclusion.

> I wonder if there is a similar system which does not actually use any 
> particular input system. Like a plain c library, I mean, where you 
> feed it bytes and it pops out unicode. 

The problem is how to handle all unicode characters with about 110 keys 
and ten
fingers. In addition, believe or not, people using CJK aren't 
unconditionally
happy with unicode because of the fact that different characters are often
assigned to a single code point.  In short, the direct use of a library you
think of would be utterly painful.

> <snip: explanation how to convert Latin characters to a Chinese one>

That's almost similar (but not the same) to what I do everyday.  So I 
understand
what you want to say...

> for Chinese, at least, I don't see why this has to be tied to X11, or 
> any other display system, it can be generic. 

Yes, it's inherently generic. In fact, any well-designed conversion server,
which is responsible for converting a sequence of Latin characters to an 
exotic
character, is independent of any underlying window system.  You can use 
it for
any non-GUI applications using its interface directly.

However, what if you have to write particular code for each conversion 
server
whenever you create a *true* internationalized application?  Awful. XIM 
makes
your life much easier; with an *unified* interface provided by XIM, you 
don't
need to care about differences between various conversion servers.

The drawback of XIM is that XIM is inevitably bound to the locale in use.
IOW, all it realizes is a *bi*lingual environment, not a *multi*lingual one.
In Unix-like environments, we have to tweak environmental variables such as
LOCALE, LC_ALL, LC_CTYPE, and others if necessrary, in order to get an
internationalized application to work properly.

The proposed text input architecture provides the user a locale-indepenent
text input environment; with appropriate input servers, you can type in any
characters of any languages without setting up locale variables each time
you switch a language to another.  Only clicking on
"Edit->Input->specific input server" on the main menu does everything 
for you.

> GNUstep itself may be in the best position to handle the actual input 
> and conversion.

I wish it is...(though I'm not sure the implication of the word "actual").

Regards,
- Kazunobu Kuriyama