TCP uses a mechanism called slow start
. This starts out transmitting information at a slow rate, and gradually speeds up until it reaches the maximum the network can handle. For long-lived connections this is a really good approach, but for small requests such as a web page, it will probably transmit the entire file in the slow phase. If you then kill that TCP connection and start another for the next file, then the next file goes through the same slow start process.
HTTP 1.1 will re-use a single TCP connection to receive multiple files one after another - but that is still slow since the server has to finish sending one file before getting told what the next file is, and then going and loading that file from storage into memory and prepping it for sending. It would be quicker if you could say 'send me the following 20 files' at the start, and then the server could keep busy by loading the next file while the first one is being sent.