Oh cool, I’ve used solve space a few times. My personal idea for how I’d engineer for that is to reframe the problem as “the geometry is a program in a program

Oh cool, I’ve used solve space a few times. My personal idea for how I’d engineer for that is to reframe the problem as “the geometry is a program in a programming language”, and have ways to calculate surface Normals, derivatives, etc. But you are absolutely correct that choice of representation / data structure makes a huge difference in what’s easy to calculate. And perhaps the best approach would be to internally have tools that can map between different data structure representations based on the calculations on the calculation at hand