Re: Q: something like autoload for coding-systems?

[Richard Stallman]

    > Depending on the patterns of coverage, other data types such as
    > bitmaps and run-length-encodings of bitmaps could provide more dense
    > representations, especially if they can be used for just parts of the
    > code space.

    It uses less space, but, isn't it time-consuming to check if
    a specific character is included or not with such an format?

Yes, it might be.

    No.  I checked it for Big5, GB2312, and CNS.  Less than 10%
    of chinese characters belongs to all of them, even if we
    take care only BMP of Unicode.

Oh well, it was worth trying.

One further idea: all this info about how to encode
Chinese chars could be autoloaded when Emacs encounters
a character in that general range for the first time.