window configurations no longer (since 21.5) include windows ! ?

[Stephen J. Turnbull]

Michael Sperber writes:

 > The thing about identity preservation pops up first here:

 > http://osdir.com/ml/emacs.xemacs.patches/2002-11/msg00099.html

Yikes!  Man, that's evil.  Totally counter-intuitive.  "We had to
destroy the window to save it."  All that just to move window-point?!?

Mike, this seriously needs rethinking.  It's just not reasonable to
invalidate window IDs across save-window-excursion under current ways
of thinking.  I agree that the `(setq outwin (display-buffer outbuf))'
idiom is probably going to be available most of the time but it's
ugly, and it's not obvious that such idioms will always work.  I'm not
even sure why it's supposed to work in this case: it seems obscure.

 > Without support from the explicit documentation, I would argue that
 > those assumptions were always unsafe.

I wish our documentation were that good.  I also think that in the
absence of an explicit disclaimer (eg. `delete-extent''s docstring
says "The extent cannot be used thereafter"), a user may reasonably
expect that a Lisp object won't spontaneously become meaningless.

Anyway, IMO Section 38.3 as quoted is sufficiently explicit.  Windows
are persistent objects and they can be revived.