Le dim. 4 oct. 2020 à 19:00, Alfred M. Szmidt <ams@gnu.org> a écrit :
       The exact same arguments could apply to C-c being "replaced" by C-x C-c.

    There is a big difference in that C-c is already used by Emacs for
    other things, one is for users, and the second is for local
    keybindings in modes.  Changing that would break quite a bit of
    things.

   I meant when C-c was chosen, I wasn't suggesting to change things now.

I suspect that C-c was choosen long before GNU Emacs.

That does not necessarily make the then-reasoning invalid. 

 But if C-x C-c for C-c was never too confusing, C-x C-z for C-z
   should be fine too.

That makes no sense to me, C-x C-c and C-c <foo> have nothing in
common, one kills Emacs by asking the user the other is used by other
modes and users.  C-c (SIGINT) in Unix has a entierly different
behaviour than all of those...

Only for programs which don't intercept it (either the key or the signal). Others choose to intercept the key or the signal and request confirmation, ask for saving, etc. 

For practical purposes, in a responsive process, C-x C-c is "quit now", aka C-c anywhere else. Even if the mechanism is somewhat different, it's not an entirely different behavior. 

I don't have a computer at hand to test, but I wouldn't be too surprised if sending SIGINT to an emacs process called save-buffers-kill-emacs.


   In any case, C-z can display a message guiding the user to the
   proper key sequence. And one could argue that C-c could show such a
   message too(without breaking anything).

C-c doesn't interrupt Emacs; so having it show a message would be
missleading to the user.

I don't understand. What would be misleading in a message "To quit emacs, type C-x C-c"?