ANONYMOUS wrote:
> 1.The design offers a good balance because 1 in 2^16 is still very low probability and a longer checksum would add to more overhead data addition
Actually Q5.
Both of your points are correct, though 1:2^16 is not really a very low probability (consider 10s of thousands of frames per second), and there is a better answer than the 2 points you've provided.
> 9. The network layer only cares about sending the packet to the right destination,and errors in data's payload are taken care by other layers
YES (taken care of by other layers, iff those layers care about it).
> 10. (the timeout part) we need to design a timeout to wait for until NACK is received correctly and if the timer is finished then sender knows it works correctly.
Negative acknowledgments will be required when (expected) data does not arrive at the receiver, and when corrupt or out of sequence data arrives at the receiver. Timeouts will be required in the receiver if (expected) data does not appear to arrive fast enough - hopefully highlighting that some data has been lost, or a return NACK has been lost.
Hope this helps,