If it works, I can get a nice boost of speed out of this algorithm.
However, if it works, I have to implement it for *real*. Which will be tricky.
Looks like it works. Geez does it ever work. Factor-of-ten reduction on some data sets. Yowch, that one was factor-of-twenty. And *that* one was thirty. None so far that have been worse than factor-of-two.
Yeah. It's worth it.
Time to get coding. (Again.)