otherwise

[David Chapman <Zvona at MIT-OZ>]

It seems bogus to me that I can't write :otherwise in a selectq.  

I think that :otherwise should be required.  There are reasonable
arguments against this.  However, I can't see arguments against
:otherwise being *allowed*.  If the idea is that it doesn't matter what
package the ``magic keyword'' (what are these things called, anyway?) is
in, then that should extend to the keyword package, too.  Unless we are
moving toward the view that keywords aren't really symbols.

Speaking of which:  Why are keywords bound to themselves?  Why not
rather have EVAL and the compiler special-case them as self-quoting?
(Presumably the compiler already does this for T and NIL.)  Surely this
would be less overhead.