Any penetration of solar wind plasma into the magnetosphere would be precluded if the plasma were strictly frozen in the magnetic field. Such transfers cannot therefore be modeled in the frame of ideal MHD: they demand nonideal effects and are likely to involve reconnection at the magnetopause. The strong magnetic fluctuations observed at this boundary have been suggested for a long time to be responsible for these phenomena. The present study revisits two questions: what is the origin of the strong magnetopause fluctuations and by which mechanisms can they allow for reconnection? It is first confirmed that the preexisting magnetosheath fluctuations can be the primary cause of the strong magnetopause fluctuations. The phenomenon invoked in a preceding paper to explain their amplification, known as Alfven resonance, is ruled out and shown to be an artifact of ideal MHD. The amplification at the boundary is instead explained by a nonresonant mode conversion (due to the magnetopause gradient), followed by a trapping of the resulting Alfven wave in the boundary (due to the magnetic field rotation). The trapped Alfven wave has strong amplitude and its finite frequency is responsible for a departure from ideal MHD associated with reconnection distributed all over the magnetopause surface. We evidence that reconnected magnetic flux, driven by the incident magnetosheath waves, is able in this way to penetrate the magnetosphere, and we show how a local reconnection rate can be estimated. This result should be the starting point for a new approach of the magnetopause transfer problem since, in this scenario, reconnection occurs without external electrostatic electric field, as in stationary X point models, and without any local instability, as in tearing type models. ¿ 2001 American Geophysical Union