The tl;dr of the article is to define handling of invalid input, so that all conforming implementations will handle it in the same way, without having to reverse-engeneer eachother to be interoperable.
So you're saying that every time I find a STEP file written in an invalid fashion, I should convene an ISO 10303 committee and wait for years to find out how everyone should handle it? That's doubly insane, because it would take many bugs that can be fixed in a day and make my customers suffer from them for years, while at the same time requiring me to modify my program to handle every bug found by every STEP software vendor or cease to be conforming.