My professor has asked this (exam) question and it actually makes no sense to me:
TCP offers a reliable, bidirectional byte stream over the best effort packet delivery in IP. For reliability, a sender in a TCP connection re-transmits a possibly lost packet. One of the methods with which TCP guesses the loss of a packet is by duplicated (repeated) acknowledgements. Illustrate clearly and succinctly this method, with a reason why it is a reasonable way to guess a packet loss.
Surely a packet is only acknowledged when it is received? If so, how can repeated acknowledgement detect a lost packet if that packet is lost and has not been acknowledged. Unless of corse he means a lost acknowledgement packet? Where that is duplicated several times to signify that it has not be received and the packet has not been acknowledged? Or is there more to this?