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

Chapter 55
Subnet Restrictions

Subnets are good allowing for a more efficient use of the address bits, but when using a routing update protocol such as RIP version 1, you must be careful about assigning a subnet mask. This protocol only allows you to assign one mask per network number. Subnet masks allows for efficiency of address space, but there are possible problems. Under a restriction of one subnet mask per network, ID can still cause inefficiencies. For example, a serial line (a telephone connection) between two sites needs only two host IDs. But with the restriction of only one subnet mask, we will still not make great use of all the bits. Under this circumstance, we would have subnet down to two bits to make the most efficient use of the address (we only need two hosts). But this will not allow us to use the address for host assignment on the LAN (unless we only have two hosts on the LAN). As you will see later, the best option is to allow variable–length subnet masks. In other words, move the mask around on different subnets that have different requirements. This is good, but you must make sure that the routing protocol (RIP, RIPv2, OSPF, etc.) understands this as well. Point blank, RIP does not, but RIPv2 does. OSPF does. Why? Routing updates have the subnet mask included in the update (it is in the link–state advertisement for OSPF). RIP does not include any subnet masks for routing entries in its table.

When using the RIPv1 routing protocol (explained later), the subnet mask must remain the same throughout a single Class B assignment. For example, if the network assignment is 130.1.0.0 and the subnet mask assigned is 255.255.255.0, the subnet mask must remain the same throughout the 130.1.0.0 network. If the network address changes (for example, to 131.1.0.0), the subnet mask may also change for this new network number.

Subnet Restrictions

RIP version 2 and OSPF do not have this restriction because they broadcast their subnet masks in the table with the network IDs (more on this in a moment).