I've skimmed the article, but haven't read the whole thing in detail. At any rate, Setterfield's claims are nonsense. I looked at them many years ago, soon after they came out. My conclusions were pretty similar to
those of Gerald Aardsma of ICR. Specifically, I noticed that:
1) Setterfield did NOT weight his fit based on the error-bars of the measurements. This is a major error, especially when the error-bars vary by orders of magnitude, as they do here.
2) Setterfield eliminated a few historical data points which were LOWER than the currently accepted value of c. This biased his fit.
3) The mathematical function which Setterfield chose for his fit had no a-priori theoretical justification. It is unusual and was chosen arbitrarily. Why not a quadratic or exponential fit, which would be more natural?
What Setterfield noticed is something very normal and natural, as pointed out by Aardsma. Early measurements of fundamental constants are often in error, and it takes time for the measurements to settle to their true values. For an example, see these
historical plots of particle properties.