> This bench shows (run with only 4 cores) that spawning is faster than > "asynccing" when generating responses requires some work.
I'm not sure where the discrepancy comes from with the 20 less QPS, but I would note that httpbeast seems to be transferring more data which may be putting the benchmark in favour of GuildenStern. To be honest, looking at the code I was surprised to see GuildenStern use `selectors`. So this is very similar to HttpBeast's approach, the main difference appears to be that GuildenStern spawns a thread per request using `threadpool` whereas HttpBeast spawns a thread per core and then runs an event loop in each. I would still consider this to be "asynccing" because utilising `selectors` is utilising async IO. I'd be interested to see a more traditional benchmark, such as the one used in the TechEmpower web framework benchmarks. :)
