Geoge Michael, he probably discovered it for good man.
He was like the best, at everything. I can still remember, his fingers dancing on the keyboard whilst his sax corner cliped the sky.
Always, when I recall that very moment, it gives me the chills.
I don't believe it was Mister or ISM. The trick has been around for a while before that post (they just figured out the specifics of it).
My guess is that it was some random person in the earlier days of SMW TASing who ran into it by accident. Like a lot of glitches, people probably didn't start finding a way to replicate it or make a use for it until the later days of TASing.
That's all assumptions, though, but I wouldn't be surprised if the case actually is like that. Nobody keeps any records of who found what glitch, though, so it could have even been found not long after the game was released.
It's not that hard to discover the corner clip glitch, when you investigate the hitbox properties of the blocks in SMW. It's not really a square, but something like this:
Actually, that's not at all how it works. ^^;
All basic hitboxes are in fact square, as the SNES wasn't capable of much else. Instead, the fault is more on Mario's fault, who's hitbox is made up of several small interaction points. It's because of the position of these points that Mario can get pushed around by corners. As for actually corner clipping, that happens when the corner of both a block's hitbox and one of Mario's interaction points run into each other. By the time the game realizes that Mario has actually touched the block, the point has traveled too far for the game to decide where to go, and no other point has hit the block to tell it. As a result, the game just pulls the rest of his points into the block. As for Yoshi, due to his larger size, the interaction points are even more spread out, and as such forcing one point into a block (or even sprite) is much easier.
So no, it's not something a person could figure out from hitboxes, as a block's hitbox is always a square. The math could be figured out by close observation, but somebody would have to know it exists for them to try something as complicated as that. Professional frame-by-frame time wizard. YouTube - Twitter - SMW Glitch List - SMW Randomizer
I don't disagree with your argument, because I didn't understand most of the interaction points code. They might explain why my picture works.
Let me explain: you can formalize and model the game's mechanics in several ways; one of them is the original source code.
My approach is not what the programmers imaginated, but it has identical "pragmatic" results. I can consider Mario as a single point and the other things as more complex and large objects and the results will be equal.
By using cheats over relevant adresses, I can fix Mario's y position (pixel and subpixel) and analyse, for each x position, where happens. It's a mapping that covers all the block's positions.
I did it particularly with a muncher and with a cement block and they are exactly the same, except the muncher hurts you.
I wrote the boundaries of the block that causes Mario to go up, left, right and down (and when he gets stuck, dies, etc). And the corners are areas that don't do anything to Mario. It's impressive how the corners from below are much larger (delta x = 5 pixels each) than the corners from above (delta x = 2 pixels each).
This analysis explain why you should have 33 speed to walljump, 49 to corner clip and the positions where corner boosting is possible, don't you think?