Truncated normal distribution

At the last paragraph of the paper the author discuss truncated normal distributions. There are some issues regarding this part.

First an noted in #3 the paper should add some explanations on "the advantage of Multiple Inference's implementation" over scipy implementation. What is the advantage? It is faster? Is it more accurate (in this case what is the expected return value)?

Second the author claims that the package supports concave truncation sets (in contrast to scipy's implementation that accepts only convex). It is not clear how to define a concave set following the documentation and more importantly it is not clear which method is used for concave truncation sets. The exponential titling method of https://ieeexplore.ieee.org/abstract/document/7408180 is only for convex sets. Also since this package is using the method of TruncatedNormal R package that should only support rectangular sets (special case of convex sets).