Mikrotik lets you make a point-to-point style address where you specify the “local end” IP as a /32 and the “remote end” IP as the network address.
The remote end doesn’t even have to be in the same subnet as anything on your side. The Mikrotik will then create a connected route to the remote address as a /32 in the routing table.
If you want to do something slightly different, then try this method:
“loopback IP of 192.168.1.1/32” (assign to a bridge interface)
“loopback IP of 192.168.1.255/32” (assign to the same bridge)
Customer1 on ether1 has 192.168.1.2
Set ether1 arp=proxy-arp
/ip route add dst=192.168.1.2/32 gateway=ether1
Customer2 on ether2 has 192.168.1.3
set ether2 arp=proxy-arp
/ip route add dst=192.168.1.3/32 gateway=ether2
Note that ether1 and ether2 should not be bridged together or switched together - each is a standalone IP interface.
Each customer configures their IP address as a normal /24
e.g.:
customer1 = 192.168.1.2/24 gateway = 192.168.1.1
customer2 = 192.168.1.3/24 gateway = 192.168.1.1
Customer1 and customer2 can actually ping each other and communicate with no problems. In fact, if the /32 routes to each interface are being redistributed into OSPF, then there could be another router somewhere else in your topology that has the same "192.168.1.1/32 and 192.168.1.255/32 loopback IPs on it, and it has a /32 route to customer3 192.168.1.4/32 → ether1.
So long as Router1 sees the /32 route to customer3 coming from Router2, then customer1 can do a traceroute to 192.168.1.4 and they will see three IP hops to get to it!
The alternative I gave works on Cisco (in fact I discovered it using Cisco, and it also works on Mikrotik, it might work on Juniper)
This solution…
… is a Mikrotik thing. (or at least I’ve never seen something like this in Cisco)
You could probably make Cisco work with it like this:
int gig0/0.1
encapsulation dot1q 1
ip unnumbered g0/1
!
int gig0/1
ip address 10.23.0.1 255.255.255.0
!
ip route 10.22.0.1 255.255.255.255 gig0/0.1
ip route 0.0.0.0 0.0.0.0 10.22.0.1
Cisco only lets you use an interface as the next hop on non-broadcast interfaces, or for ethernet interfaces, it works if the destination interface is an ethernet sub-interface.
