It's not great that that's referenced, and the doc page in general isn't wonderful (nor was the idea of 'just exposing' something without abstracting in uAPI).
But at the same time, surely relying on you hardcoding kernel-specific value might be a clue that it's not a great interface to use?
So on one hand I agree with you the doc is bad and shouldn't have implied you can do that, but on the other hand a user relying on this should have been more than a little wary.