Hello Tom
I only wanted to sharpen the argumentation in this field. Your input test curves despite the arc are single span degree 2 and 3 Bezier curves. This means that an approximation algorithm using a single span degree 2 bezier should stop with error zero if used on a single span degree 2 Bezier curve. And your results show that even with higher degress the error does not vanish on your examples. So your algorithm is not optimal in a mathematical sense leading to the conclusion that there is a better solution with less degree/less spans possible than the one you presented.I even attached a demo where i added a low degree Bezier approx of your arc example within your tolerance claims. I doubt your ‘very good approximation algorithms’ can top this result. Give me some pointers and I’ll check them out.