I find strange the assumption that bytecode is bigger than the source. It's not true for many programs and many languages. Since we have interpretation phase a

I find strange the assumption that bytecode is bigger than the source. It's not true for many programs and many languages. Since we have interpretation phase anyway, it would be interesting to try to optimize for size and speed of interpretation or pre-compilation.