Comment by int_19h

Stable ABI and especially the issue of generics is why I think bytecode+JIT is the right approach to a language that operates at this level of abstraction. Higher-level semantics of bytecode allow for much more flexibility wrt ABI, and JIT lets you compile generic code monomorphically across ABI boundaries, and even inline it. A long time ago I did some experiments with very carefully written C# code, and it was capable of the same abstraction as C++ STL containers and algorithms, while producing highly efficient native code because everything was inlined at runtime.