> Markm wrote: > > Having NaNs be larger than all non-NaNs, including positive > > infinity, is > > surprising too. What EVER you do, don't screw with the ordering of NaN and such. Too many programmers already understand and depend on that algebra.