RE: delete-windows-on

["Drew Adams" <drew.adams@oracle.com>]

Let me summarize the current state of the discussion and my
questions/suggestions:

1. I suggested moving `delete-windows-on' to Lisp. Apparently, Martin already
plans to do this, and the performance hit is expected to be negligible.

2. In Emacs 23, arg BUFFER is optional. Prior to Emacs 23, it is required. So
the treatment of a nil BUFFER has changed. My objection was with the pre-23
approach; I was unaware of the new treatment. Dunno if this behavior change
breaks any code - we'll find out, I guess. It is a user-visible change, in any
case, so it should probably be in NEWS (it is not, so far).

3. Emacs 23 always returns nil. Dunno what the case was before 23. I don't have
a problem with it always returning nil.

I did suggest that we might instead return something indicating whether a window
was actually deleted, similarly to how `kill-buffer' lets you know whether it
killed the buffer. But this suggestion is not so important.

4. Raising an error for a non-string, non-buffer BUFFER arg is consistent with
other functions. OK, I have no problem with that.

5. Raising an error for a string that does not name an existing buffer is wrong,
IMO. No other opinions expressed about this, so far.

6. I suggested that for interactive use the completion candidates be limited to
buffers that actually have windows. No other opinions expressed about this, so
far.