Demonstration of naive and incorrect way to compare web performance.

bad_performance_test.sh

#!/usr/bin/env bash

# This script demonstrates a naive and incorrect way to compare the 
#  performance of HTTP, HTTPS (HTTP/1.1), and HTTPS (HTTP/2).

# This is incorrect because:
#   1. the HTTP response is just a 302, not a full page.
#   2. each request is just for the initial page html and 
#       not all of the resources required to render the full page.
#   2. each and every HTTPS request is a completely new 
#       TCP connection and new TLS session. 

for i in {1..100}
do
    H1=`curl -w "%{time_total}\n" -o /dev/null -s http://www.cachefly.com/`
    HS=`curl -w "%{time_total}\n" --http1.1 -o /dev/null -s https://www.cachefly.com/`
    H2=`curl -w "%{time_total}\n" --http2 -o /dev/null -s https://www.cachefly.com/`
    echo "${H1},${HS},${H2}"
done

echo ''