Even in that case, similar problem can occur. In general, a Unix
journaling
filesystem cannot completely prevent halfly written files. The problem
is
due to Unix concurrency semantics (things written are immediately visible
by
all other processes, so if in the middle the processes all dies there is
no
safe point to get back to). If you use a filesystem that looks like that
of
Ameoba or Sprite, that's another matter (but, let's get back to reality).
So in general, a journaling filesystem is good in that you won't trash
your
filesystem. But that's all it can guarantee to you. Perhaps the average
amount of failure can be reduced, but it is not guaranteed. You can
still
need your backup copy of available, or a UPS.