Ok, you made me honestly curious. What makes you think writing a tokenizer in C++ is impossible ? I didn't see any deadline, but is it because of time ? They even provide some algorithmic hints on how to do it...

The project aims to build not just the tokenizer, but a complete compiler and standard library. None of GCC, Clang, Microsoft or Intel can provide a complete, bug free C++11 implementation. If you hang out in freenode's ##c++, you'll see people finding new corner cases every day. It's reasonable to consider this project impossible for an individual.

A tokenizer is very reasonable. I would even have believed a complete C compiler, although that is a huge undertaking. But C++ is, as far as I am aware, the hardest language to compile in popular use.

The tokenizer is the easy part. The overall goal of this, writing a C++11 compliant compiler + associated standard libraries is going to be near impossible.

I think he means the full course goal, given that the first assignment assignment is not the hard part.