On an aside, relating to the performance of these two Firefox technologies, it is interesting that in the battle of the JavaScript engines it is artificial metrics that are driving the game: No one is measuring the number of clocks taken to run the JavaScript on typical pages like Twitter or gmail, or whether the variance is enough to even noticeably matter, but instead we run looping benchmark code entirely unlike anything actually used in practice (SunSpider being the most egregiously invalid test).

I would posit, based upon nothing, that in practice JIT systems can be a net negative to the overwhelming majority of run-once JavaScript encountered across the tubes.