Add reference implementation for higher numbers of concurrent requests to HTTP1Server about swift-nio

[helje5]

Your "end" counter is flawed, you need to put your counter in the schedule block:

Not this:

         case .end(_):
+            queue.sync {
+                if concurrentRequests > maxConcurrentRequests {
+                    maxConcurrentRequests = concurrentRequests
+                    print(maxConcurrentRequests)
+                }
+                concurrentRequests -= 1
+            }
+            
             _ = ctx.eventLoop.scheduleTask(in: delay) { () -> Void in

but this:

         case .end(_):
             _ = ctx.eventLoop.scheduleTask(in: delay) { () -> Void in
+              queue.sync {
+                  if concurrentRequests > maxConcurrentRequests {
+                      maxConcurrentRequests = concurrentRequests
+                      print(maxConcurrentRequests)
+                  }
+                  concurrentRequests -= 1
+              }
+            

(you count before the artificial "delay")

[GeorgeLyon]

Whoops, I actually did catch this but posted an incorrect gist (I have updated it). This just adds 1. to the max (before it was 5, after it becomes 6) which is consistent with my theory that a single request seems to be blocking the next request from coming in.

[helje5]

How do you "spam the server"?

[GeorgeLyon]

Ah thank you! You guessed correctly. My test environment was going between HTTP2 (for reference) and HTTP1 (for swift-nio, since it doesn't support 2 yet). Maximum number of HTTP1 connections (for my client) was 6. Increasing the maximum worked correctly.