Performance doesn't exist in a vacuum. The fastest way to optimise a loop is to remove all the code in it, but you're changing the program in order to do so. A

Performance doesn't exist in a vacuum. The fastest way to optimise a loop is to remove all the code in it, but you're changing the program in order to do so. Adding/removing IO buffering absolutely _does_ have side effects. If someone is running `./app >> /dev/null 2>&1`, then mucking arouind with IO buffering is going to have little to no effect.