Because the older, classful routing protocols don't convey prefix lengths, routers using these protocols have no way of knowing the prefix lengths of incoming routes! Therefore, they have no choice but to assume that subnet routes from different parts of their classful network all use the same subnet mask that they use. This leads to a further restriction on subnet design - the same subnet mask must be used throughout the classful network, a practice called Fixed Length Subnet Masking (FLSM), as opposed to Variable Length Subnet Masking (VLSM), supported by the newer, classless breed of routing protocols.

For example, assume that we've assigned a 210.22.75.128/26 subnet. If we're using RIP (an older, classful, FLSM routing protocol), we can not now assign 210.22.75.96/28 as another subnet. Why? Because both subnets are in the same class C network (210.22.75.0), yet they have different prefix lengths. FLSM forbids this. We could assign 210.22.75.64/26, however, since both subnets in the 210.22.75.0 network would share the same prefix length of 26.