In view of some of what you wrote above, I'm compelled to note that Cook is generally well-informed and his posts are on target - as is this one. I think if you read through some of his posts you'd agree. Some of the ad hominem comments you made above are regrettable and baseless.

(About my own experience with floating point: I think I wrote my first Cholesky solver in K&R C in about 1986, while taking a course based on Golub and van Loan. This was before IEEE was sensibly implemented and you could observe significant FP differences across various Unix platforms.)