How is Riemann–Stieltjes Integration insufficient for developing modern probability theory?

288 Views Asked by At

If we consider Riemann–Stieltjes integration then it can perfectly account for mixed probability distribution (a continuous R.V with some point mass). So why would we still need Lebesgue Integration theory?

Is it because the Riemann integrable class is not large enough, or is it because under Riemann integration interchanging limits and integration is too hard(usually requiring uniform convergence)?