Morphing Attack, i.e. the deception of Face Recognition Systems (FRS) through a face morphing process between the identity of two subjects with criminal intent, has recently emerged as a serious security threat. Due to its significance, recently several Morphing Attack Detection (MAD) systems, i.e. methods based on Artificial Intelligence able to automatically detect the presence of morphing, have been proposed in the literature. Unfortunately, developing, comparing, and reproducing these MAD algorithms is challenging, particularly for deep learning-based solutions, since they are usually evaluated on private datasets and the source code is not publicly released. Therefore, we observe the need for an open-source framework that aims to simplify the development of new MAD systems, in combination with their evaluation. Thus, in this paper, after a discussion about the current limits of existing studies on the MAD task, we examine the desired properties and features of this framework, with a particular focus on its modularity, usability, and effectiveness.