I in part reproduce this very interesting paper for the sake of a remark which appears to me important. I write (a, b, c, f, g, h) in place of Muir's (A, B, C, A′, B′, C′), and take as usual (A, B, C, F, G, H) and K to denote (bc − f2, ca − g2, ab − h2, gh − af, hf − bg, fg − ch) and the discriminant abc − af2 − bg2 − ch2 + 2fgh.