Self-organizing peer-to-peer (p2p) overlay networks like CAN, Chord, Pastry and Tapestry (also called distributed hash tables or DHTs) offer a novel platform for a variety of scalable and decentralized distributed applications. These systems provide efficient and fault-tolerant routing, object location, and load balancing within a self-organizing overlay network. One important aspect of these systems is how they exploit network proximity in the underlying Internet. Three basic approaches have been proposed to exploit network proximity in DHTs, geographic layout, proximity routing and proximity neighbor selection. In this position paper, we briefly discuss the three approaches, contrast their strengths and shortcomings, and consider their applicability in the different DHT routing protocols. We conclude that proximity neighbor selection, when used in DHTs with prefix-based routing like Pastry and Tapestry, is highly effective and appears to dominate the other approaches.