How can benchmarks validate the wrong opinion just because it is popular, when they are benchmarking time and memory use?