Some rules

There are some branches, given the nature of a relation, are going to fail - idea is to prune them using a cut.

(1)  relation(...) :- ... .
(2)  relation(...) :- ... .
...
(i)  relation(...) :-
       t1,t2,t3,!,t4,t5,t6.

(i+1) relation(...) :- ... .
(n) relation(...) :- ... .

• Relation has n clauses (in order indicated)
• Relation (i) examined if relation (1) to (i-1) fail
• t1,t2,t3 fails - the cut doesn't take place so instance (i+1) of relation is chosen
• If t1,t2,t3 succeeds, the cut takes place and computation proceeds to prove t4,t5,t6
• if t4,t5,t6 fails, because of the cut, there is backtracking on t1,t2,t3, instance (i) of relation fails and instances (i+1) to (n) of relation are ignored (not considered as possible candidates to unification for a subgoal)