The paper presents a model for designing wormhole routing algorithms that are deadlock-free, livelock-free, minimal or nonminimal, and maximally adaptive. The model does not rely on adding physical or virtual channels to network topologies but instead analyzes the directions in which packets can turn and the cycles that these turns can form. By prohibiting just enough turns to break all cycles, the model produces routing algorithms that are deadlock-free, livelock-free, and maximally adaptive. The paper applies this model to n-dimensional meshes and k-cubes, describing specific routing algorithms for each. Simulations show that the partially adaptive routing algorithms perform better than nonadaptive algorithms for nonuniform traffic patterns, particularly at high throughputs. The paper concludes by discussing the advantages and disadvantages of the turn model and suggests future work directions, including extending the model to other network topologies and identifying realistic workload distributions.The paper presents a model for designing wormhole routing algorithms that are deadlock-free, livelock-free, minimal or nonminimal, and maximally adaptive. The model does not rely on adding physical or virtual channels to network topologies but instead analyzes the directions in which packets can turn and the cycles that these turns can form. By prohibiting just enough turns to break all cycles, the model produces routing algorithms that are deadlock-free, livelock-free, and maximally adaptive. The paper applies this model to n-dimensional meshes and k-cubes, describing specific routing algorithms for each. Simulations show that the partially adaptive routing algorithms perform better than nonadaptive algorithms for nonuniform traffic patterns, particularly at high throughputs. The paper concludes by discussing the advantages and disadvantages of the turn model and suggests future work directions, including extending the model to other network topologies and identifying realistic workload distributions.