It was stated that the extra chainwheel should be of a size size such that the increase of the ratio accomplished by the second chainwheel was as close to half the increase accomplished by moving from one rear cog to the next. Astute observers noticed that this assertion is technically incorrect. What you really want to do is to make the actual ratio of the large chain ring to the small as close as possible to the square root of the actual ratio of successive ratios which would be obtained by the rear cogs alone.

In practice, it makes precious little difference. In the case of a 16% increase at the rear, we are talking about the difference between 1.08 and sqrt(1.16), or ~1.077. Since all actual ratios obtainable are limited by the discrete properties of chain drive, or, indeed, toothed wheel transmissions in general, making such a fine distinction is impossible (indeed, maintaining the constant ratio of ratios obtained from the rear cluster can only be approximated, and is not always exactly what you want to try to do anyway!). The difference between halving the increase, and taking the square root of the ratios could, however, influence precisely which discrete choice of chainring size you did make.

On the other hand, I get the impression that, because of modern marketing practices, the choice of chainring and cog sizes is even more discrete now than it was in times past.