How can a Data Transfer Protocol using NAKs only be reliable?

Question

I have been studying a book on Computer Networking (this is not a homework question)

One of the questions compares ACK and NAK based data transfer protocols, with the emphasis being that for a NAK based protocol, packet loss can for packet x can be detected by the received when the (x + 1)th packet is received.

However, my issue is that what happens if the NAK sent by the receiver is lost before it gets to the sender? The sender will not be aware of errors and will not retransmit. Furthermore, what if the packet is the last one in the sequence? (there is no subsequent packet to test with)

I do not see how a NAK-only protocol can be reliable (delivers each packet in the correct order)

Answer 1

I suspect that the theoretical context described in the book assumes an infinite stream of packets and ignores conditions like the channel becoming completely disabled. In practice, there are other constructs defined as part of the ACK/NAK protocol or left up to higher or lower layer layer protocols to deal with. For instance, a physical layer would provide indications like "the channel is broken", and perhaps a protocol above sends/receives "end-of-stream" packets, with a timer detect the case where the NAK for the last packet is lost. These are just hypothetical examples, but are the types of things real-life protocols do.

I do not see how a NAK-only protocol can be reliable (delivers each packet in the correct order)

I'm sure the book is implying there is some sort of mechanism for identifying packets in order. Otherwise, there would be no way for the receiver to indicate which packet it is NAK'ing, i.e., most protocols use a sequence number in each packet/ack/nak.

Answer 2

A NAK-only protocol can achieve reliability by having the receiver start a timer when sending a NAK and retransmit the NAK when the timer times out before the missing packet was retransmitted.