TechEmpower
No more fakes
Please, no more http: length hard encoded.
No more template response cache !!! The template is OK to cache, but the response for the fortunes is NOT OK.
https://github.com/TechEmpower/FrameworkBenchmarks/pull/9560#issuecomment-2607279905
mrhttp also non realistic implementation: https://github.com/TechEmpower/FrameworkBenchmarks/issues/9055
Now the devs of the framework decide if it's realistic or not.
We need to implement more tests before start the benchmark !!
https://github.com/TechEmpower/FrameworkBenchmarks/issues/9440#issuecomment-2518472173
The template is OK to cache, but the response for the fortunes is NOT OK.
I wonder if we should make fortunes response dynamic to prevent this. Maybe store more fortunes in the database and pick random fortunes, similar to the query test?
Shouldn't pre-computed HTTP headers trigger "Stripped" classification? #8205
@p8 https://github.com/TechEmpower/FrameworkBenchmarks/issues/9578#issuecomment-2656955552 https://github.com/TechEmpower/FrameworkBenchmarks/issues/6529 https://github.com/TechEmpower/FrameworkBenchmarks/issues/6788 https://github.com/TechEmpower/FrameworkBenchmarks/issues/6883
The aspnetcore framework precomputes the header length as well:
https://github.com/TechEmpower/FrameworkBenchmarks/blob/c82a834c5e8aeae14c8203cc05beaaa18dcc7cfc/frameworks/CSharp/aspnetcore-mono/PlatformBenchmarks/BenchmarkApplication.Json.cs#L14-L18
https://github.com/TechEmpower/FrameworkBenchmarks/blob/c82a834c5e8aeae14c8203cc05beaaa18dcc7cfc/frameworks/CSharp/aspnetcore-mono/PlatformBenchmarks/BenchmarkApplication.Json.cs#L12
https://github.com/TechEmpower/FrameworkBenchmarks/blob/c82a834c5e8aeae14c8203cc05beaaa18dcc7cfc/frameworks/CSharp/aspnetcore/src/Platform/BenchmarkApplication.Plaintext.cs#L13-L16
aspnetcore
image
https://github.com/TechEmpower/FrameworkBenchmarks/blob/c82a834c5e8aeae14c8203cc05beaaa18dcc7cfc/frameworks/CSharp/aspnetcore/src/Platform/BenchmarkApplication.Json.cs#L19-L23
The benchmark uses a custom implementation of the Date header generation:
https://github.com/TechEmpower/FrameworkBenchmarks/blob/c82a834c5e8aeae14c8203cc05beaaa18dcc7cfc/frameworks/CSharp/aspnetcore/src/Platform/BenchmarkApplication.Json.cs#L25-L30
But in fact, we should test the implementation that comes with the original framework:
https://github.com/dotnet/aspnetcore/blob/main/src/Servers/Kestrel/Core/src/Internal/Http/DateHeaderValueManager.cs
https://github.com/dotnet/aspnetcore/blob/8645583a97fa8140da0a588bbd00c5e9cf13983c/src/Servers/Kestrel/Core/src/Internal/Http/HttpProtocol.cs#L1263-L1269
Please, no more http: length hard encoded.
No more template response cache !!! The template is OK to cache, but the response for the fortunes is NOT OK.
It will be really nice
Unfortunately, it will be a massive reduction in a number of approved frameworks. A lot of solutions just use dirty hacks to show higher values.
People sometimes are protecting this position with "Nginx-like proxy will be a protection layer and our server with logic can do whatever we want and we don't care about proper HTTP implementation"
If you need a Nginx proxy to present a proper HTTP implementation, then the Nginx proxy needs to be in the test (welcome to the bottom of the charts, lol).
