The Online Action Detection (OAD) problem needs to be revisited. Unlike traditional offline action detection approaches, where the evaluation metrics are clear and well established, in the OAD setting we find very few works and no consensus on the evaluation protocols to be used. In this talk we propose to rethink the OAD scenario, clearly defining the problem itself and the main characteristics that the models which are considered online must comply with. We will also present some novel metrics, which have been specifically designed to evaluate models online. Finally, we will show the latest advances in unsupervised systems for the generation of action proposals in an online video stream.