CS wise, performance is pretty much garbage in NS no matter how you write the code and you typically aren't dealing w/ large enough datasets that performance matters that much (there are cases where you do and want to take a closer look).. but nobody is digging down to big O/little O/theta etc..