You made great points, and I've realized that the compiler/interpreter doesn't always know what's the part of the context that we are actually interested in. Th

You made great points, and I've realized that the compiler/interpreter doesn't always know what's the part of the context that we are actually interested in. There's often too much context available and it's not always easy to throw away the information that we're not interested in. However when one does manage to make a function to a structural representation that contains exactly the interesting things, then equality is this function followed by structural equality.