Were C++20 coroutines the best approach for stackless coroutines?

I keep looking at that paper and wondering about the advantages and disadvantages of the adopted approach vs the one in the paper.

Does anyone have more insights?

- The paper uses the `resumable` keyword. Meaning that you have to use `resumable` in each function prototype when no deduction can be done.

- You have to mark as `resumable` the evaluation mode.

- There is no viral `co_await` all the chain up.

- There is `resumable` deduction. This makes the standard library templates, for example, reusable and convertible to `resumable` by injecting your `resumable` function/iterator/whatever.

- It seems to be easier (correct me if I am wrong) to calculate the space needed for a coroutine?

- If follows something much closer to a function-object approach.