One-hot encoding vs. n-ary decision trees

Hi,

With one-hot encoding decision trees remain binary but their depth increases.
What about extending the algorithm to handle node splits among more than 2 branches?

Thanks,

Laurent.