Multicast is not scalable mainly due to the number of forwarding states and control overhead required to maintain trees. Tree aggregation reduces the number of multicast forwarding states and the tree maintenance overhead by allowing several multicast groups to share the same delivery tree. In this paper, we exhibit several drawbacks of the existing protocols: the latency to manage group dynamics is high, the managers are critical points of failures and some group-specific entries are stored unnecessarily. Then, we propose a new distributed protocol that significantly reduces the number of control messages and limits the number of trees within a domain. By simulations, we show that our protocol achieves good performance and outperforms the previous known distributed algorithm.