enter image description here The image attached was included in a question where I was required to design the network f0r the classful network address, and it says that each circle represents a subnetwork. My question is why would we consider the connection between 2 routers as a network? Does a connection between 2 routers even need an IP address?

  • 1
    Typically in a point to point connection you will always need two IP addresses, one on each physical interface. The subnet used could for instance be a /30.
    – user36472
    Nov 6, 2018 at 11:26

2 Answers 2


Just a comment about terminology and how it colours our view.

"Subnet" is a word which really suggests there are classful, "full", networks which are divided into portions called subnets. (See for example the diagram on p6 of RFC 950 "Internet Standard Subnetting Procedure"). This hasn't been real life for decades.

The current classless mechanism (RFC 4632) just recognises the network portion and the host portion of the address. We don't have subnetworks, just networks of various sizes.

With this modern view, the network with just two things on it is merely a small network.

What is the internet? It's an "interconnected network of networks" -- the normal definition of "internetting" (wiki). So pretty much all networks are connected at their edges. Your example drawing is four networks joined together by two routers. It happens that there are no endpoint hosts on one of them.

The dominance of the internet has meant that we think of connecting our things to the internet -- inherently an assymetric view of the world. Technologically though, what we actually have are small networks (an office, say) joining a big internetwork by making a new two-router network in between.

Does a connection between 2 routers even need an IP address?

No, you can run IP point-to-point links unnumbered, and many people do. It's a little known fact that in principle, routers don't need any IP addresses on any interfaces. This is because when a host forwards to a next-hop router, it sends the packet inside a suitable layer-2 frame such as ethernet or PPP, and that packet has the ultimate target's IP address as destination. The IP address of the next-hop router was only used to find the layer-2 address. In practice, however, IP routers always have at least one address, so you can connect to it for configuration, and (as far as I know) there are no operating systems for hosts or routers which permit ethernet addresses as the address of the next-hop router. A secondary reason is that ICMP packets from the router (eg time exceeded etc) normally come from the nearside address.


Forward traffic by gateway IP address is router's function, every network(such as VLAN) has gateway IP address on router(or third layer switches) , the router every port has different network's IP.

And the two routers border to each other should in one network, then they can communicate, because they are transfer data in layer 2 by MAC address as source and target address, router is the 3rd layer device, between two routers there is no device for forwarding networking traffic.

  • "every network(such as VLAN) has gateway IP address" ... that is an oversimplification I would dare to call dangerous. There are countless IP networks that have no gateway of any form, because the design does not want them to. For example private inter-host networks that should never become reachable from anywhere else at all - sometimes even implemented on separated switches. IP networks they are - routers or gateways, they have not. Nov 6, 2018 at 17:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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