In the world of modern surface design tools, a mesh is simply a data structure – a scaffold for higher-order basis functions which specify the “true” surface. The price paid for this analytic continuity is complexity in seemingly-simple tasks such as refinement, composition, and direct manipulation. But consider the lowly triangle mesh. With the ability to split and merge edges wherever more detail is needed, or stitch and sew arbitrary boundaries, unstructured high-resolution triangle meshes bring us much closer to the ideal of modeling at the level of surfaces rather than data structures (as has been previously noted).