Let $f \in C_b(S)$ (set of all bounded and continuous functions) and $\mu$ be a measure on $S$ where $S$ is a complete separable metric space. Then a book (Probability Theory by Borkar) claims that for a given $\epsilon > 0$ there exists $N \geq 1$ and $a_0 < a_1 < a_2<\dots <a_N$ such that
(a) $\|f\| -1 = a_0 < a_1 < a_2<\dots <a_N = \|f\| + 1$
(b) $\mu(\{x | f(x) = a_i\})=0$ for all $i$ and
(c) $a_i - a_{i-1} \leq \epsilon$, $1 \leq i \leq N$ .
I am not understanding how can he claim point (b) for a general $\mu$.
Then he claims $\bigcup_{i=1}^{N} \{x|a_{i-1} \leq f(x) \lt a_i\}=S$.
$\|f\| = \sup_x|f(x)|$