>> Are you volunteering to write an fsdb for every file system which is
>> ever incorporated into the source tree that supports hard links to
>> directories?  8-)
> If that's what it takes, yes.

Or to put it another way, an fsdb analog has to be considered an
essential part of implementing a new filesystem, as important as an
fsck analog.  I agree with this point of view: unlink() should never do
anything to a directory.  If there is some twisted reason to want to
make it succeed, it should do the same thing rmdir() does.

More generally, no filesystem syscall should be able to turn a clean
filesystem into a filesystem that needs to be fscked.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu