RE: (special-mode-map): Delete binding for `z'.

[Drew Adams]

> What is the rationale for this change?

Not needed?  Error prone?  Could be used for something more useful/appropriate?

This was the reason RMS gave (bug thread #12311):

 I just discovered why some of the Dired buffers I always keep
 occasionally disappear.  It is because the binding of z in
 special-mode-map makes it so easy to kill them.

 I just got rid of that binding in my .emacs, but I think that
 binding is a bad idea.  Not all the buffers in special modes
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 are things people don't mind deleting, and this key can cause
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 annoyance if typed by accident.

> I use this all the time in buffers like *Help*, 

Why bother to kill *Help*?  Why do that rather than `q'?

> *Dired* &c.  What am I supposed to use instead? C-u q?

Sure, why not?  Or `q' if you don't need to kill it.
Or `C-x v' if you want to replace it.

You want to kill Dired but you don't want to use `C-u q'.
OK, so bind `z' to killing once again if you want to.

Or is it about other special-mode buffers?  That's the useful question, to me.
I would say forget about *Help* (which doesn't need to be killed) and Dired
(which you sometimes don't want to kill).

But are there special-mode buffers where `z' killing the buffer is really
useful?  If so, then maybe the fix was too heavy-handed.

> How is a ding more valuable than a useful behavior?

Killing the buffer is not the only or the most useful behavior `z' could have in
Dired.  Dired is a rich mode with lots of keys already taken and with plenty of
commands that could be bound to keys.

Whether `z' should be bound to killing the buffer in `special-mode-map' is a
different question.  RMS's fix was to remove that binding.

I argued instead that Dired should not inherit from `special-mode-map' - it
should just define the few keys from `s-m-m' that it needs.

The question really is whether the binding of `z' makes sense for special modes
in general (e.g., beyond *Help* and Dired).  RMS apparently thinks not.  But the
problem (annoyance) he reported was really about Dired.

[Followup should really be in the bug thread, BTW.
Moving it here means people can miss that discussion:
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=12311.]