Suppose $f$ is a non-negative Riemann integrable function in $[a,b]$. Is this true that $$ \sup_P \sum_{j=1}^{n} |f(c_j)(x_j-x_{j-1})| = \int_{a}^{b} |f(x)|dx$$
where $c_j \in [x_{j-1}, x_j]$. I don't think so. But in the definition of bounded variation it is used http://en.wikipedia.org/wiki/Total_variation#Basic_properties
$$ \sup_P \sum_{j=1}^{n} |f(c_j)|\cdot(x_j-x_{j-1}) = (b-a)\cdot\sup_{[a,b]}|f|$$ Proof: For every positive $\varepsilon$, consider a partition with $(x_0,x_1,x_2)=(a,u,b)$, $|f(u)|\geqslant\sup\limits_{[a,b]}|f|-\varepsilon$, and $c_1=c_2=u$.
The only case when such $u$ do not exist is when the supremum is reached at $a$ or $b$, then a subdivision into the unique interval $(a,b)$ does the job. QED.
Is it? Where?