So this information theory

It is about finding the spot on a binomial where entropy is maximizing, the term where noise deviations partition signal deviations with the greatest integer ratio.

Shannon says this deviation ratio has to meet Shannon Nyquist, self sampling at sustainable sampling rate.  It has to count to integers, and thus a digit system in its minimization. This is the reason for the log, bandwidth is an exponent rotating about the flat unit circle.

For any given message set, Shannon will have some surplus samples, and floors to integer set, will be some connected nodes on the Markov triples. Surfaces of color operators.

But looking at the total number of samples as a contiguous empty elements, Then any given set of similar samples will meet the -iLog(i), as that selects the set's share of sample space where the entropy is maximizing. This is also redundancy minimizing.  This is the point on the binomial where that partition ratio is maximum.

Part one was just selecting the N, total number of samples to be partitioned. That is what Shannon-Nyquist was about.  The extension of the theory is the complete self sampled system where mean, variance and skew are matched between these message sets. The concept of relative sampling and sampling at twice the bandwidth are retained in the Markov n-tuples. The 'Gaussian' condition is met by allocating on the -iLog(i) basis, part two.

In Gaussian terms the theory says that for any given signal to noise there is a point of large N, number of samples, where each message set has a sufficiently accurate binomial to fit the tail end of a Gaussian distribution.  The is the N to infinity thing on flat earth.