Accurately measuring per-hop packet dynamics on an Internet path is difficult. Currently available techniques have many well-known limitations that can make it difficult to accurately measure per-hop packet dynamics. Much of the difficulty of per-hop measurement is due to the lack of protocol support available to measure an Internet path on a per-hop basis. This thesis classifies common weaknesses and describes a protocol for per-hop measurement of Internet packet dynamics, known as the IP Measurement Protocol, or IPMP. With IPMP, a specially formed probe packet collects information from intermediate routers on the packet's dynamics as the packet is forwarded. This information includes an IP address from the interface that received the packet, a timestamp that records when the packet was received, and a counter that records the arrival order of echo packets belonging to the same flow. Probing a path with IPMP allows the topology of the path to be directly determined, and for direct measurement of per-hop behaviours such as queueing delay, jitter, reordering, and loss. This is useful in many operational situations, as well as for researchers in characterising Internet behaviour. IPMP's design goals of being tightly constrained and easy to implement are tested by building implementations in hardware and software. Implementations of IPMP presented in this thesis show that an IPMP measurement probe can be processed in hardware without delaying the packet, and processed in software with little overhead. This thesis presents IPMP-based measurement techniques for measuring per-hop packet delay, jitter, loss, reordering, and capacity that are more robust, require less probes to be sent, and are potentially more accurate and convenient than corresponding measurement techniques that do not use IPMP.