Those familiar with Windows NT 4.0 will remember that by adding more that one network card to a system and enabling IP forwarding, you could use Windows NT as a router. Though the functionality was limited to acting as a static router or one which could only exchange information with other routers using RIP version 1, the ability to have NT act as a router was often used where a hardware-based solution (such as a Cisco router or similar) was impractical or too expensive. Windows 2000 builds on this functionality, with the Routing and Remote Access service (RRAS) providing the ability to integrate with other routers using a variety of popular routing protocols including RIP versions 1 and 2, as well as OSPF. Further to this, RRAS will also allow your server to act as a demand-dial router, initiating dial-up connections (as well as VPN connections) via ISDN and standard phone lines. This demand-dial functionality provides what could potentially be a very cost-effective solution in offices where Internet or related dial-up costs (such as WAN connection) are prohibitively expensive.
Before having a discussion about configuring a router, I think it is first important to understand what a router actually does, especially besides the obvious (routing packets). For the sake of simplicity, lets consider a 2-subnet internet. In order for hosts on one subnet (who have a given address range) to talk to computers on another subnet, they must communicate using a router as an intermediary. Sometimes referred to as a gateway, the router has a connection on both networks, usually with separate network interface cards, one on each subnet. When a host on one subnet needs to talk to a host on another, it forwards the frame it has created to the local router interface. Upon receiving the frame, the router does a number of things. First, it strips off the associated frame addressing (for example the Ethernet MAC addresses), and then looks at the destination IP address. Though the router (usually) won’t know about the whereabouts of a specific host, it will know about the networks to whom it is attached at a minimum, as well as any it has learned about via routing protocols. If the router has the destination network in its routing table, it will note the IP address to where the datagram should be sent next, either the destination host itself, or another router (if applicable). After decrementing the TTL of the datagram by 1 (as happens at every router), the router them frames the datagram for the underlying network technology, including the appropriate MAC addressing, and forwards the frame to that host.
Whenever you talk about routers you should be sure to distinguish between routing protocols and routed protocols. Quite simply, a routed protocol is one whose traffic has an addressing scheme that allows it to be routed, such as IP or IPX. On the other hand, a routing protocol is one that routers use to exchange information with one another, such as RIP or OSPF.