other-notes

policy: until a piece is assembled, only download sub-pieces for that piece
TCP pipelining: should have multiple pieces pending at once
every time a sub-piece arrives, send new request

connection start: select random piece and start 

send have message: after 'piece' message
if we get a have message: send request message if we don't have that piece
on bitfield message, request random piece (block) we don't have


logic: have all peer threads decide on 1 piece to download
then each thread requests blocks inthat piece, move on once completed

thread pool:
20 threads
initialize -> each thread connects to a peer
distribute work -> client decides to work on 1 piece
                -> peers each given 1 sub-piece (block) to work on
                -> if their peer doesn't have it, return & wait
                -> if they do, download it, then return & wait
                -> returned threads continue getting subblocks
                -> once piece is done, re-assign threads to new piece