Except where comparisons being made, the first and second halves are mostly independent. If the tree-walking interpreter might not be what you have hope for, you can, in principle, skip finishing it. The bytecode VM half is a re-implementation (in C instead of Java as the first half) and only the scanner/tokenizer has some inevitable redundancies in implementation. For me, who had less prior knowledge in interpreters, it is still good to have fully gone through both halves to relate to the comparisons and trade-offs of both implementations.