Hi Pavel Nedelko, that’s a valid point.

The reason why the tests lack any real logic is to test the minimum overhead you get from running said framework and not test external libraries that parse request bodies, or fetch data, etc. This overhead applies to every request, whether it’s hello world, or database fetching, or whatever else.

So if the minimum overhead is on average 3ms more, a slow request that takes 300ms, will be 303ms and so forth.

Remember that it applies to every request, throughput will definitely go down, even if we disregard the extra latency on every request.

Does the latency itself matter? Well that’s the point of debate. For example, minifying HTML, the savings can be minor. And that savings only applies to requests that serve HTML, versus on every request.

Though, I think more than the performance difference I just wanted to point out that implementation matters, that frameworks don’t need to inherently incur overhead by nature. That any of the frameworks I mentioned could’ve been done with the same or similar features and be faster or slower, and that it really boils down to implementation.