Illustrated TCP/IP by Matthew G. Naugle Wiley Computer Publishing, John Wiley & Sons, Inc. ISBN: 0471196568   Pub Date: 11/01/98

Chapter 94
Indirect Routing

If the source and destination stations are on different networks, they must use the indirect routing services of a router. The transmitting station will address the physical destination address of the packet to that of the router (using ARP, if necessary, to find the physical address of the router) and submit the packet to the router. Each workstation may be able to determine the address of its closest router, or it can be preconfigured with the address of its default router. The router has two choices:

1.  If the destination network indicated by the address in the IP header is directly attached to the router, it will forward the packet directly to the destination station.

2.  If the destination network indicated by the address in the IP header is not directly attached to the router, it must use the services of another router to forward the packet and let that router determine the next hop path.

Notice here, the destination physical address is that of the router and not the final destination station. This type of routing is indirect routing. The final destination IP address is embedded in the IP header.

Sending a packet to its final destination might be accomplished by using both direct and indirect routing. For example, when a packet is to be delivered across an internet, the originating station will address it to the router for delivery to its final network. This is indirect routing. The originator and destination may be separated by more than one router. No matter whether the final destination network ID is directly connected to that router or whether the packet must traverse a few routers to reach its final destination, the last router in the path must use direct routing to deliver the packet to its destination host.

Depending on the Options field settings, it should be noted that the original IP datagram, will not be altered with two primary exceptions: the TTL (Time to Live) field and the Cyclic Redundancy Check field. If an IP datagram is received by a router and it has not arrived at its final destination, the router will decrement the TTL field. If TTL > 0, it will forward the packet based on routing table information. The IP datagram’s header contents will remain the same (with the exception of an error-detection field known as the Cyclic Redundancy Check, or CRC). Since the TTL field changed, the CRC must be recalculated throughout all the networks and routers that the datagram traverses. Otherwise, the only alterations that are made are to the datalink headers and trailers. The IP addresses in the IP header will remain the same, as the datagram traverses any routers in the path to its destination.

IP routers forward datagrams on a connectionless basis and therefore do not guarantee delivery of any packet. They operate at the network layer, which provides best-effort or connectionless data transfer. Routers do not establish sessions with other routers on the internet. In fact, IP routers do not know of any workstations (nonrouters) on their subnets.

These routers forward packets based on the network address of the packet (in the IP header) and not on the physical address (the 48-bit address for broadcast networks) of the final destination (the receiver). When the router receives the packet, it will look at the final network address (embedded in the IP header of the packet) and determine how to route the packet. Routers only route packets that are directly addressed to them. They do not operate in promiscuous mode (watching all LAN traffic) for forwarding datagrams.