0

The Wikipedia (and other sources) page about the OSI model explains layer 2 as such:

The data link layer provides node-to-node data transfer—a link between two directly connected nodes. ...

Let's take Ethernet (802.3) for example, which is a layer 2 technology (as well as layer 1, but that's not of our concern for this question). In an Ethernet network (LAN), hosts communicate using MAC addresses through a switch (possibly), so we have an important functionality of addressing here. Why isn't this a part of the definition of the layer 2 functionality?

1
  • Ethernet is also not the only layer-2, 'tho it is very popular these days. (Frame relay, and ATM aren't totally dead.)
    – Ricky
    Sep 11 '21 at 20:23
2

The data link layer provides node-to-node data transfer—a link between two directly connected nodes.

That isn't correct - nodes don't need to be directly connected. It's the purpose of a layer-2 protocol to enable point-to-multipoint communication. Each node can reach any other node by simply sending a frame to the destination's MAC address.

You might have quoted a definition for a physical layer connection which is almost always point-to-point nowadays.

we have an important functionality of addressing here. Why isn't this a part of the definition of the layer 2 functionality?

But it is. You seem to have read the wrong texts or may have misunderstood what you read.

Ethernet covers OSI layer 1 and 2. Originally, all nodes were connected by a shared wire and each NIC filtered all frames not addressed to its own MAC address. The interconnecting network was just passively distributing everything everywhere. The actual addressing was all done on the end nodes. When the wire was replaced by a multiport repeater (hub) that logic didn't change.

Later, switches came up that replaced the essentially passive network by controlled, directed and buffered forwarding. That way, the efficiency of a network was greatly improved without needing to change any logic at the nodes/NICs.

4
  • So the wrong part of this definition is "directly connected" ?
    – YoavKlein
    Sep 11 '21 at 20:06
  • Most layer-1 protocols are point-to-point today, so directly connecting nodes limits your network to exactly two nodes. Using concentrators (switches, obsolete repeater hubs), your network can connect large numbers of nodes.
    – Zac67
    Sep 11 '21 at 20:15
  • 1
    the wrong part is "two"
    – Effie
    Sep 11 '21 at 20:18
  • 1
    For ethernet switches, yes, you still have a p-t-p construct. Host talks to switch, switch talks to hosts. The switch does what is necessary to get your traffic where it should go -- unicast to the known dest port, flooded to all ports (broadcast, unknown dest), or copied to the correct ports (multicast).
    – Ricky
    Sep 11 '21 at 20:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.