In general, all the people who read this paper have not a clear idea of the optimizations.

It's clear that the idea of consistent that the authors have is not quite correct.

Why comparing copy on write with journaling file systems?