Perhaps by compiling from the bytecode of the existing interpreter, instead writing a whole new source code interpreter for the JVM, it will have some benefits in supporting newer versions of the language. Or, maybe it will help to decouple development of the standard library and the interpreter, possibly letting more devs work on it. You can't really be sure what benefits/deficiencies a different approach will bring until you actually try it - see PyPy, a "silly" idea to implement Python in Python, which now yields faster runtime performance than CPython.