Maximizing entropy on a channel

Question

In Shannon's paper "A Mathematical Theory of Communication", in Theorem 8 he states:

Theorem 8: Let the system of constraints considered as channel have a capacity $C = \log W$. If we assign $$p^{(s)}_{ij} = \frac{B_j}{B_i}W^{\ell^{(s)}_{ij}}$$ where $l^{(s)}_{ij}$ is the duration of the $s^{th}$ symbol leading from state i to state j and $B_i$ satisfy $$B_i = \Sigma_{s,j}B_jW^{\ell^{(s)}_{ij}}$$ then H is maximized and equal to C.

My question is, what exactly is $B_i$ and $B_j$? It's not defined anywhere else. In a different section he mentions "a sequence $B_i$ of symbols from the source", but that doesn't make much sense here.

Answer 1

Charles E.Radke, "Necessary and sufficient conditions on conditional probabilities to maximize entropy", Information and Control 9, 279-284 (1966), https://doi.org/10.1016/S0019-9958(66)90165-3 gives a proof of Shannon's Theorem 8:

A necessary and sufficient condition such that $\eta_c$ is maximized and equal to $1$ is that the set of symbol lengths ${l_{ij}}^v$ must satisfy: $${p_{ij}}^v = (\beta_j/\beta_i)\omega_r^{-{l_{ij}}^v},\tag{3}$$ where $\beta_i$ and $\beta_j$ are respectively the $(k, i)$th and $(k, j)$th cofactor of the matrix $A$ for any $k$.

The matrix $A$ is defined earlier as:

$$A \triangleq \left[\textstyle\sum\omega^{-{l_{ij}}^v} - \delta_{ij}\right];$$

which is something you can recognize in Shannon's Theorem 1.