I created 2 empty files in VS Code, one with UTF-8 encoding(file.utf8.txt) and the other with GB 2312 encoding(file.gb2312.txt).

If we use the od command to view the contents of these 2 empty files, we will find that the character encoding used by a file does not exist in the file itself (text files do not have so-called headers to store metadata). This looks like no problem.

$ touch file.utf8.txt
$ touch file.gb2312.txt
$ od -t x1 file.utf8.txt
$ od -t x1 file.gb2312.txt

Next, I start up the Chinese input method(IME) and use the same Latin letter sequence “zhongwen” to input 2 Chinese characters “中文” in these 2 files, separately. Regardless of which file, the Chinese characters can be displayed normally without glitches or mojibake.

However, when I use the od command to view the contents of the 2 files again, I find that the IME seems to produce different outputs for the same input “zhongwen” as follows,

$ od -t x1 file.gb2312.txt
0000000    d6  d0  ce  c4                                                
0000004

$ od -t x1 file.utf8.txt  
0000000    e4  b8  ad  e6  96  87                                        
0000006

I wonder if the IME can “telemeter” (with some Inter-Process Communication tricks) which encoding the editor is currently using and perform some kind of output conversion, or does the IME always produce the same output for the same input, such as always outputting the Unicode encoding of the 2 Chinese characters “中文” regardless of the current encoding of the editor, and then the operating system (or other magical components) converts this output into UTF-8 or GB 2312 correspondingly?