Some TCP software constructs the destination Ethernet address from the top 24 bits of the Ethernet address followed by the low 24 bits of the (class A) destination Internet address. For this scheme the top 24 bits of the Ethernet address must be the same on all hosts on the net. Contrast ARP.
Try this search on Wikipedia, OneLook, Google
Nearby terms: constant folding « Constantine/Yourdon « constant linear velocity « constant mapping » constraint » constraint functional programming » Constraint Handling In Prolog