We informally introduce here the information flow-control domain. We then present two theoretical layouts for developping programming language with information flow-control. The approach is semantic-based and allows programs to be statically type-checked. We then consider its practical use, through the so-called Paragon language, extending Java in a user-friendly way. We’ll be especially intersted in issues about its compiler, written in Haskell. Specifically, we’ll take a close look at both the problem of infering non-specified policies, and the performance of a null-pointer analysis.