It is much more interesting to know the performance model of a given language than comparing languages. If you know the performance model, you know what costs,

It is much more interesting to know the performance model of a given language than comparing languages. If you know the performance model, you know what costs, so you begin writing programs that avoid doing expensive things. And it will always be the case that the clever algorithm or data structure is much better at getting some cycles back than switching language. Part of the article is also wrong. A dynamic lookup can often be avoided if you have a dataflow analysis phase in the compiler for instance. It is almost "the" optimization for OO languages like Java.