Docking at the data link layer, we are each frame of cyclic redundancy detection (CRC), the content of the CRC for the whole frame are tested, and the probability of error is very small, so we approximate think all passes through the CRC test frame is correct, in the same way, the IP packet is placed within the frame, may be divided into multiple packages placed in multiple frames, so since the CRC check frame correctly, also means that IP packets also have no mistakes, then when the data link layer frame part of data submitted to the IP layer, IP layer directly to forward it to go, why want to use first and test is to check? Because the solution is the bits inside the package was not wrong question, since the CRC is done, why IP to do, and apparently first and inspection check the effect is not good CRC, (because I think the data bit bit error will appear this kind of situation only happen in the signal transmission over long distances, can only be seen as CPU dispatch data from memory to the Cache will not to check, I think in the internal integration equipment, data transmission is not wrong, so the data link layer is passed to the IP layer data will never go wrong, at this point I think the first IP test and is totally useless, if there is wrong, CRC has long been found, if yes, also not solve the substantive issues)
Reliable transmission need to be solved: bit gone wrong, frame order arrived, repeated frames, lost frames these problems
CRC can be thought to have solved the bit error problem, then the IP protocol, the transport layer protocol should focus on solving the frame order arrived, repeated frames, lost frames a few questions, so don't have to completely in IP layer first and inspection,
Have a saying: the IP layer does not consider the frame order arrived, repeated frames, lost frame these issues, these will make the routers in the network is very complex, increase cost, so simply focus on IP layer routing, the packet check completely to the data link layer in the hand down to do it,
In short: I don't think CRC has to ensure that the IP packet first will go wrong, so in the test first and it's meaningless,
Sincerely hope that each ace to give directions, thank you,

CodePudding user response:

Data link layer not only Ethernet,
I need to change and it is the historical reason, also not line

CodePudding user response:

CRC can be understood as a compression, although the probability is very small, but there are possible: there are multiple bytes changed at the same time, calculate the CRC is the same as the right, to detect the just the way they are

CodePudding user response:

But the final CRC result or more bytes at the same time changed the correct probability is too small, generally we seem not to think about it,
If the frame is correct, so even if the IP packet is divided, its division also are right, in other words, together is also correct, IP just guarantee packet forwarding correctly, the reliability of the other problems don't go to tube, so why should first and testing?

CodePudding user response:

Providing the error control at the link layer is an optimization, never a requirement.

TCP/IP network layer and transport layer, and the realization of the data link layer has different

CodePudding user response:

This is because networking is an agreement between parties, that at the layers, we have the company standards committees, and because there is politics says the and rivalry between those committees that prevents logical and beneficial outcomes.

CodePudding user response:

I see, thank you for your advice ~