The Smith-Waterman algorithm returns the single best local alignment,
but two proteins may share more than one common region. Waterman
and Eggert [32] have shown how all local alignments may be obtained for
a pair of sequences with minimal recalculation. Recently, Barton [33]
has described how for a simple length dependent gap-penalty, all locally
optimal alignments may be determined in the order of steps without recalculation.