Q:
2. "ARP protocol does not reside in any part of layer 3"
I guess this is the crux of the debate. I hope to show you that it does.
3. "ARP protocol is used only in collision domains....a layer 2 environment."
I believe you mean broadcast domain here, otherwise how would a node on one switch port resolve IP to MAC of another node
on the same switch but connected to a different switch port as the two switch ports would be in two different "collision
domains"?
4. "ARP protocol is only used on interfaces to resolve IP to MAC."
I don't think this is true either. ARP is used to resolve protocol addresses to hardware addresses (in the case of the earlier
conversation this happened to be IP). RARP is used to resolve hardware, hardware or MAC addresses to protocol addresses (
in the case of the earlier conversation IP), and Inverse ARP is used to resolve DLCI numbers to IP addresses. Now granted
RARP and IARP are differnet implementations of ARP but they are still ARP. There is also AppleTalk ARP as well. Not sure
about IPX, but I am sure there is an ARP implementation for that as well.
5. "ARP protocol does not cross subnets by default."
I never said it did.
6. "ARP protocol is used to facilitate the transfer of frames only in a collision domain."
In my opinion this is incorrect, as I stated earlier this would mean that two nodes on the same switch connected to two
different switch ports would then not be able to translate IP to MAC of the alternate node. ARP itself only resolves protocol
addresses, IP in this case, to hardware addresses and then places them into the ARP table. This would occur, for our
conversation, in a broadcast domain. It is the broadcast nature of Ethernet that facilitates the transfer of frames.
7. Regarding your comments on the presentation of RFC 826.
|