Re: XIM String Conversion Callback

[Mark Leisher <mleisher crl nmsu edu>]

    Steve> I think there are likely to be contention problems here between
    Steve> what the app thinks is the current context, and what the IME might
    Steve> do. Defining context requires a model of "context", and that varies
    Steve> between input methods, even for a single language. Is the context
    Steve> the last few characters which form a polysyllabic word in Chinese?
    Steve> Is it the sentance to date (where "sentance" can be a slightly
    Steve> vague term in some languages)? The XIM documentation doesn't really
    Steve> address these issues. It just vaguely refers to "context".

The "::retrieve_context" signal would provide the amount of text before and
after the cursor that would be considered the context.  Since this signal is
likely to be emitted by input methods (assuming I understand Owen's idea),
they should know how much text constitutes "context."

I do think this should be done in terms of characters and not bytes just to
avoid counting problems that might arise from different input method and
application encodings.

In the case of Chinese, sophisticated input methods that "guess" the next
Hanzi from prior input would need a varying amount of context to do it right.
Pinyin (and maybe BoPoMoFo) input would also need variable size context to
handle automatic spelling corrections.

In short, only the input method really knows what it needs.
-----------------------------------------------------------------------------
Mark Leisher
Computing Research Lab            Cinema, radio, television, magazines are a
New Mexico State University       school of inattention: people look without
Box 30001, Dept. 3CRL             seeing, listen without hearing.
Las Cruces, NM  88003                            -- Robert Bresson