Mesh routing operates in the physical and data link layers to allow packets to flow through a dynamic mesh using multiple hops. We previously talked about mesh-under and route-over routing, but we will go somewhat deeper in this section on that form of routing.  

6LoWPAN mesh networks utilize two schemes for routing:

• Mesh-under network: In a mesh-under topology, routing is transparent and assumes a single IP subnet representing the entirety of the mesh. A message is broadcast in a single domain and is sent to all devices in the mesh. As previously mentioned, this generates considerable traffic. Mesh-under routing will move from hop to hop in the mesh but only forward packets up to layer two (data link layer) of the stack. 802.15.4 handles all the routing for each hop in layer two.
• Route-over network: In a route-over topology, networks will incur the charge of forwarding packets up to layer three (network layer) of the stack. Route-over schemes manage routes at an IP level. Each hop represents one IP router. 

The following graphic depicts the difference between mesh-under and route-over routing:

A route-over network implies that every router node is equally capable and can perform the larger set of functions as a normal IP router, such as duplicate address detection. RFC6550 formally defines the route-over protocol RPL (ripple). The advantage of the route-over architecture is the similarity with traditional TCP/IP communication. RPL provides multipoint-to-point communication (where traffic from devices in a mesh communicate to a central server on the internet) and point-to-multipoint communication (central service to the devices in the mesh).

The RPL protocol has two modes for managing the route tables:

• Storing mode: All devices configured as routers in a 6LoWPAN mesh maintain routing and neighbor tables.  
• Non-storing mode: Only a single device, such as the edge router, maintains the routing and neighbor tables. To transmit data from one host to another in a 6LoWPAN mesh, the data is sent up to the router where its route is calculated then transmitted to the receiver.

The routing table, as the name implies, contains the mesh routing paths, while the neighbor table maintains each node's directly connected neighbors. This implies that the edge router will always be referenced to deliver packets in the mesh. This allows the router nodes the freedom of not managing large routing tables to manage, but does add some latency in moving packets since the edge router must be referenced. Storing mode systems will have higher processing and memory requirements to manage routing tables stored on each node, but will have a more efficient path to establishing a route.

Note in the following figure the hop count, source address, and destination address fields. These fields are used during the address resolution and routing phase. The hop count is set to an initial high value and then decremented each time the packet propagates from node to node in the mesh. The intent is when the hop limit reaches zero, the packet is dropped and lost from the mesh. This provides a way to prevent runaway networks in the case a host node removes itself from the mesh and is no longer reachable. The source and destination address are 802.15.4 addresses and may be in short or extended format, as 802.15.4 allows. The header is constructed as: