> Emacs's handling of files tends to be fairly vague about how it handles
> file identity, so I think the fine distinction between "cp+rm" and "mv"
> in this case can be ignored.
I would have to disagree here. At the very least, as a *user*, when I
look for a command to rename a file, if I find a command that copies
and deletes the original, then I'm going to keep looking. Also,
renaming a file is atomic on many filesystems, whereas copy+delete is
certainly not going to be. Why would we want the "official" way to
rename a file be needlessly prone to race conditions, even if the
problem doesn't come up in practice?