I agree that pressing cardinal directions is a different universe than the games were written for. I agree that it's not an oversight of a developer. But here we have to deal with the reality that already exists, and while not being okay with it, we still always need to investigate it to find the internal flaws if they exist.
The concept that exists now for almost 10 years and that no one of the actual TASers disagreed with so far is:
- Game code can not be modified, it is taken as is.
- Controls the game can handle are all considered legal.
I mention that not because I think so personally (though I do), it is just how it is for all these years. I now also see that there is a huge similarity between using the buttons not present on the primary controller and by using the ones that are present but can't be pressed at once. The legality of these 2 cases seems equal to me. But there are 2 ways to merit this very legality.
1.1. Abstraction says that nothing can be taken with absolute seriousness, since then it will always contradict something else. And if 2 different people take 2 contrary things with absolute seriousness, they will always argue hard with no constructive result. So sometimes we need to shut our eyes to some minor improbability to achieve something much more enjoyable and important.
For example, the game determinism as we completely rely on here does not exist in reality. "TASing became feasible because of general assumption that behavior of a game is always determined by its initial state and player's input, and nothing else". In fact, the game behavior can depend also on how much dust covers a disc, what is the room temperature, and whatever else. There are enough factors that can break the console syncability, games can behave differently even from load to load with all the similar timing. But it all has no way to be accounted, so our emulators only account for the initial state and user input, and if they match, the game does the same as well. This is where perfect emulation starts contradicting perfect determinism. At TASVideos we need both hard, so we can't actually achieve both equally perfect. Some will need to be sacrificed (accuracy most likely).
1.2. Abstraction is critical for actual users (TASers), because they need something to rely on completely before they start serious work. Note how many problems people have with sync unstable emulators. Determinism is what provides them a comfortable environment for productive work. Accuracy is a bonus we can afford as emulation is being developed.
It is also a paradigm of creating TASes. We must abstract from the internal game rules as much as possible, because then we could find the tricks that are improbably to discover playing by the intended rules. And there are 2 ways to make the game do something: hacking its memory/code (illegal) and sending it controls it can handle (legal). TASers are in fact dealing with the stripped game code and ignore a good portion of the rest.
There always can appear a TAS that refuses to use simultaneous presses, reset+saving, zipping, whatever. It would always have a right to be published if the viewers like it. But not to degree of banning the use of what it refuses to use. All are happy.
2. Technically, there are official Nintendo controllers that allow pressing any amount of buttons at once, and have more buttons than the primary controllers:
http://en.wikipedia.org/wiki/Power_Pad
So when banning some aspect we may easily discover it is actually not worth banning and is acceptable. And here abstraction once again pops up: nothing can be considered 100% perfect.