A good path tracker is one of the keys for the successful development of a self-driving car. In the literature, there exists a wide variety of techniques, some complex and some simple and yet effective in particular scenarios. The choice of the path tracker influences the performance in terms of precision, stability and passenger comfort. This paper addresses the lateral control of a self-driving car in an urban environment, where speed is not high but variations in velocity and curvature are frequent. In choosing a lateral controller, simplicity, efficiency and robustness are considered as the main criteria. In this paper, three classical techniques used for controlling the lateral error are analyzed: pure pursuit, Stanley and a simplified kinematic steering control. Additionally , a novel kinematic controller based on the lateral speed is proposed. A home-made realistic simulation environment has been developed to allow rapid testing of the control laws. The relevance of this work has been demonstrated for all controllers through realistic simulations and experiments. The experimental site is the campus of Ecole Centrale de Nantes, where all control laws have been compared along the same path. A longer path, involving a portion of the ring road of Nantes (France) has been simulated. It involves speeds up to 90 km/h, allowing to extrapolate the comparison results to higher velocities.