>However, in my opinion, this lack of validation (really, lack of parsing in any meaningful sense) means that the performance between this parser and Needletail is incomparable. So what exactly does it demonstrate? You get to claim your implementation is faster than someone else if you do the same task in less time, but not if you skip half the job.<p>Yeah, this makes the original post just incredibly misleading. Sure, meaningful benchmarks are hard, but before you go writing "woah, look how much faster we are!", a bit more due dilligence should be required.