Sorry David :-)  But remember:  Yours was slow, but at least it worked!
 Mine was slow, and didn't even work right....

I would NEVER dispute the authority of 'benchmark' but running Lyndons
vs mine, and even Sylv's scripts (on the bash command line), shows it
to be consistently quite a bit slower (10X slower) on a 10,000 element
array, w/ 2,000 dup's.  What's up with that?  I think on the big array,
I got a 'thin' Zero... Lyndon got a 'fat' Zero!