On Thu, May 03, 2007 at 08:37:40PM -0400, Greg Troxel wrote:
> "James K. Lowden" <jklowden@schemamania.org> writes:
> 
> > One of these days, I'd like to do/see a presentation called "Journey of a
> > Keystroke" that walked through all the layers of mediation and
> > interpretation, and how they're controlled.  I'd like to know why --
> > sooner or later, with enough indirection via ssh/rsh -- the backspace key
> > starts being echoed as ^H when I mistype a search in less(1).
> 
> Basically:
> 
>   The IBM PC put a Backspace key where the DEL key ought to be, or maybe
>   this is from IBM 3270 etc.
> 
>   To atone for their sins, they made BS act like DEL acted from V7 unix
>   on (and also on DEC operating systems).
> 
>   There are two ways to cope with this (meaning the odd placement of BS,
>   together with software like emacs that expects DEL):
> 
>     1) Map Backspace to DEL, making the key do what the key that's there
>        should.  Don't change anything else, meaning stty erase ^?.  This
>        is my approach.  In xterm, Control-Backspace is DEL normally, and
>        if you set *ptyInitialErase to true, Control-Backspace is then ^H
>        (with plain Backspace DEL).

That's how the PC (PS/2) keyboard work under wscons. USB keyboard is different
for some reason. (PR kern/32755, PR kern/32754 )

As the PC keyboard works fine as it is now, I propose to change the USB
keyboard mapping to match the PC keyboard. As you note below, doing it the
other way seems too complicated.

>     2) Leave Backspace as ^H, but then stty erase ^H.  This sets up for
>        a whole cascading kludge where many things are changed to cope
>        with this:
> 
>        a) emacs has code (by experimentation) to remap characters if
>        erase is ^H so that pushing the key where DEL ought to be does
>        what you expect.  It does this by remapping BS to DEL before the
>        keymap sees it.
> 
>        In this mode, it's hard to get help.  Even worse, some things (I
>        can't remember anymore) don't get remapped properly.
> 
>        Hint: Don't try C-M-BS if you are running X.   I have done that
>        *twice* while experimenting and composing this message....
> 
>        b) When sshing to a remote system, and maybe telnet, propagate
>        the value of stty erase.  Logging in from NetBSD to NetBSD with
>        OpenSSH does this.  So the programs on the far side also remap ^H
>        to DEL.
> 
>        If you log in to a remote system in a way that doesn't do this,
>        you end up with a mess.  This can happen with a serial port, or a
>        non-NetBSD system.
> 
> 
> So, I'm in favor of fixing the problem once by fixing the keymap.

Pavel