I find that some of the above comments do not incorporate the fundamental thesis in the original article (that I have not read!) which is the issue of Exit Pupil. In case people are not clear as to what this is, let me put it this way. The only light (-cone)that matters in focusing manually is what gets into your eye, not what the lens collects or what the optical train projects. If the pupil of your eye in the ambient illumination is, say, 2mm diameter, but the Exit Pupil of the camera optics (lens+Fresnel screen+viewfinder optics) is 7mm, then your eye will simply not be receiving the full effective aperture of the imaging lens. You will see a DoF that applies to a much smaller f-stop than the nominal value of the f-stop of the lens. The write-up quoted states that the Exit Pupil of the viewfinder system, including the Fresnel screen is such that only around f4 actually comes out of the back end of the viewfinder, not the f1.2 of the imaging lens, never mind what your pupil has opened up to. Hence, with such a viewfinder system it would be impossible to fully appreciate the shallowness of the DoF of fast lenses with your eye, and your focusing will consequently be more sloppy than it should be. IMHO this argument would apply to both film cameras and DSLRs and the only way to focus such a fast lens properly would be to use Live-View, as is now available in the Canon 40D.
Last time I tested the autofocusing in my now ancient 1Ds (for astrophotography) I discovered that the autofocus was extremely reliable with lenses f2.8 or slower, but considerably less so with faster lenses. With a 1.4 lens there were random errors, but of course the focus was always within the DoF of f2.8. Now, in retrospect, perhaps the autofocus system of the 1Ds does have an effective exit pupil equivalent to f2.8? Food for thought.