Right, so it's entirely appropriate to benchmark arbitrary computational metrics, but benchmarking UI metrics is inconceivable?
I don't understand your argument - true shitty engineers can ruin a UI, as well as they can ruin a JS interpreter. The point of testing and benchmarking is to evaluate the relative "performance" in each of these domains.
Measuring page loads on an array of popular, representative websites, or measuring the latency between user interactions and some visible response are some examples.
Also, having a group of real people give subjective ratings I think carries more weight than people give it credit for.
I don't understand your argument - true shitty engineers can ruin a UI, as well as they can ruin a JS interpreter. The point of testing and benchmarking is to evaluate the relative "performance" in each of these domains.
Measuring page loads on an array of popular, representative websites, or measuring the latency between user interactions and some visible response are some examples.
Also, having a group of real people give subjective ratings I think carries more weight than people give it credit for.