|| > And here is a BUG-report and suggesion related to mule input: || ... || || I can't reproduce it. What I tried is: || C-x 2 ;; split window vertically || C-x b temp RET ;; switch to a new buffer "temp" || C-u C-\ chinese-array30 RET ;; activate "array30" || C-x o ;; goto the lower window || C-u C-\ chinese-zozy RET ;; activate "zozy" || C-x o ;; goto the upper window || aaax ;; type some letters || || Please show me exactly what you did like the way above. open a new Emacs session: C-u C-\ chinese-zozy RET ;; activate "zozy" o ;; enter `o', you will see a customized radical very similar to Japanese character `he' C-x 2 ;; split window vertically C-x C-f tmp.txt ;; find a blank tmp.txt file on the above window C-u C-\ chinese-array30 RET ;; activate "array30" on above window, leave the below "zozy" ;;now the above and the below windows with different files and input methods o ;; enter `o' again, you will see "9^" radical on minibuffer (so far so good) C-x o ;; change active window to the below one o ;; enter `o' once again while expecing a radical similar to Japanese "he" since the below one is using "zozy", you see a "9^" radical on minibuffer, just like the above one (using "array30".) And the bug is caught. (If you are editing the same file in two windows, there will be no bug observed, since Emacs would automatically use the same input method for one file.) Determining which input method's radicals to be shown on minibuffer seems OCCURS on ``changing input method from one to another(denoted as THIS)'' and set as THIS method's radicals to be shown. Therefore if ``determining which input method's radicals to be shown'' occurs on ``changing the active window'' and set as ``the active window's input method's radicals to be shown'', this bug might be thrown into trash bin... || > 2. Leaving a blank after a word is being inputted is a waste || ... || || I'm not sure I understand what you mean correctly, but... || || It's a very complicated feature. First of all, currently || Emacs has to lookup the whole key sequences of an input || method to know how to type a specific character in that || input method. In addition, it seems very annoying to see || such information every time one types characters. || || Even if we make such a command, I think it should be bound || to some key (e.g. C-u C-u C-x =) instread be automatically || called while using an input method. I got your points (through observing how mule input method is working.) But I believe that there must be an elegant means to respond this need for Chinese user. Anyway, I will keep thinking about this, wishing we will finally find an esay way to follow... || > I guess this feature is also useful in other languages... || || I think it's very specific to Chinese. No other languages || use such an ragical-based input methods. Once again, I feel that Chinese is really a writing language, although there is no much difference with other languages such as English to me. It seems that Chinese is unique to all other languages. Does not Japanese or Korean have input method(s) of ``spelling shapes of words''? If the answer is NO, preparing ARRAY30 for Japanese would absolutely be very interesting. :)