Please, Guys. Let's get NOT get emotional about this please! (It must be the spring air. There is even an endiannes flamewar on linux-kernel). Personally i think we have been conservative about it way too long. We really should have done it for 21.0 because we had a good reason then: The NT port (which is a good justification) and a major version number increase. The argument _against_ doing it in 21.0 (and maybe 20.4) is that the tty-erase-char stuff has not been tried in the field yet. Anyway, Steve, How come it is broken for you? This is the code that is used (defsubst delete-forward-p () (and delete-key-deletes-forward (or (not (eq (device-type) 'x)) (x-keysym-on-keyboard-sans-modifiers-p 'backspace)))) So apparently XEmacs thinks that on your keyboard Backspace can be generated without a modifier and thus that it is safe to change the keyboard. This is not a question of defaults. This looks like a real bug. Jan P.S. Let's all remember that set-mark-command is also bound to a key which can be close to impossible to generate. Can you say 'telnet.exe'?