According to examples across the web, if two routers are directly connected with an ethernet cable, this is a point-to-point network (which totally conforms to rfc point-to-point definition). but if there is a switch between routers, that makes a broadcast network. how it is concluded from rfc definition.
You can define a network by the topological configuration or by the capabilities of the underlying (data link) protocol.
From a topology perspective, two routers connected together are a point to point network, simply because there are exactly two devices on the network. A talks only to B, and B talks only to A. It wouldn't matter what the underlying protocol is (Ethernet, WiFi, DSL, etc).
From a protocol perspective, the same network could be considered a broadcast network because of the features of the data link protocol. Some protocols, like Ethernet and WiFi have broadcast capability -- that is, they can address a single message to all hosts on the network. Other protocols like HDLC (now obsolete) can't do that. They are designed to send messages to exactly one host that is on the other end of the wire.
At the time OSPF was developed, there were many more data link layer protocols in use than there are today. OSPF makes the distinction between broadcast, point-to-point, etc, because of how neighbors discover and communicate with each other, and how LSAs are exchanged.
Depending on the capabilities of the underlying data link protocol, OSPF can discover neighbors using broadcast/multicast, otherwise neighbors have to be explicitly configured.
OSPF assumes a broadcast network has more than two neighbors, so it uses the designated router concept to minimize LSA exchanges.
By selecting the network type, OSPF can make the most efficient use of the router resources by choosing the best way to talk to neighbors, minimizing bandwidth and CPU -- two things that were precious when the protocol was developed.
