trait One 
{
  fn one(&self); 
}
trait OneTwo : One 
{
  fn onetwo(&self); 
}

The code that implements OneTwo must also implement One (the same as when we overrode the default method, we still had to define is_done), therefore:

struct Three; 
impl One for Three 
{ 
  fn one(&self) 
  {
    println!("one");
  }
}
impl OneTwo for Three 
{
  fn onetwo(&self) 
  {
    println!("onetwo");
  } 
}

And the result is as follows:

If we omitted the impl One block, we would get a compilation error complaining that impl OneTwo requires impl One to exist.