also trying to understand it.
according to python:
1) you send a request
2) you MUST get response headers for (1) (THIS IS MANDATORY)
3) you send another request
4) you get response body for (2)
5) response headers for (3)
6) response body for (5)
Only two requests can be pipelined. Surely this is an unavoidable, wait no it's just written into the code to error out if you don't do it that way.
according to reality:
1) you send a request
2) you do not get response headers for (1)
3) you repeat steps 1-2 until enough responses are queued
4) you receive those responses as header,body,header,body...
they even name it with a __ so to make it hard to override, but the state can safely go to Idle after a request has sent, whether or not response headers have come in. Sure the connection might close, but then you adjust to not pipeline, and re-send the rest of your requests over a new connection.
Python http.client has pipelining disabled in the most hard coded fashion possible with python, making implementing pipelining deliberately difficult. BUT I COULD GET ALL THE PREVIEW THUMBS AS A ONE BIG BLOB NOT LEETLE PIECES
