> believing that all voting machine software source code should be open source.
That much is true, for principled reasons.
As far as his proof by contradiction--his analysis isn't actually good proof of that. If he were trying to just show that correlations should have been of slope >=1, he could have just plotted that correlation--but he didn't.
Instead he did a weird thing where he plotted this derived quantity and arbitrarily fit a piecewise linear model to it. This seems very sketchy and intentionally misleading to me.