It remains to prove that the above method produces a normalised path as its output, given that the path P1 was normalised. The proof is basically a proof by contradiction; we show that if the output path of the algorithm is not normalised, then the input path must also have not been normalised.
Given that the input is
The output of the algorithm is a path P2, with the properties as shown in fig. 3.3
Any alias for P2 is an alias for P1 with b inserted at some point. So suppose we have a P3, an alias for P1
We must have , as P1 is normalised. If P3 =
P1, then we clearly cannot insert b at a valid point in order to
decrease P3, so assume from now on that P3 > P1. This
gives us:-
Now suppose P'3 is P3 with b inserted at some point:
There are now several cases to consider depending on the relative sizes of (c,k,q). For each case we aim to show P'3 > P2, and conclude that P2 must be normalised.