> I don't believe it's a bug for the system to allow the superuser to do
> stupid or dangerous things.

...famous last words.   I think it might be reasonable to say that it
can be assumed that the superuser knows what it's doing, but that
doesn't mean that you should overload dangerous functionality on top
of standard syscalls.

If kernel support for linking and unlinking directories does belong in
the kernel (I personally agree with jtk that it doesn't), then it
belongs there as unlinkdir(2) and linkdir(2), not as an overloading of
link and unlink.

			       _MelloN_