LP relaxations interpreted as metrics (or distance functions in graphs) are useful in designing approximation algorithms for cut problems. This was illustrated by designing an exact algorithm for minimum s-t cut, a 2-approximation for multiway cut in graphs, a 2-approximation for the multicut problem in trees and finally a logarithmic approximation for the multicut problem in general graphs.