This is much more a statement about how simple the ATA port interface is than it is about Forth. Writing for hardware only sounds complicated to modern folks who have never tried it. Remember that these were devices built with a few thousand transistors at most. There's only so much complexity you can pack into that structure.
I wish there were more resources available detailing how to write drivers. I've had this dream in the back of my mind for a long time of hacking a standard USB port on the end of a Steel Battalion controller, and then writing a PC driver for it. But I wouldn't even know where to start, really.
You probably don't need to write a driver if the controller plugs into the Xbox's controller port. You can just download the Xbox controller drivers written by others.
That said, you may find the following resources useful (I certainly have):
Linux Device Drivers (a very helpful book, but the kernel API has changed a bit since its release):
It's kind of strange that they built something custom instead of just implementing it as a couple standard HID devices with an internal hub.
Depending on the USB chip/microcontroller used, it can be a lot easier to just toss in a bunch of vendor-specific requests instead of adhering to all the requirements of an HID-class device (though I haven't analyzed the Xbox controller itself, so I don't know if that's what was done).
Also, there's no need to use an internal hub, as a single USB device ("function" in the USB spec, IIRC) can have multiple "interfaces" that implement different (or the same) classes, such as HID, CDC, mass storage, or vendor-specific.
I would not call it "complete". Maybe "complete" in the sense that it works under perfect conditions, but not "complete" in any real sense. I don't see any error handling or power-up initialization, I don't see any DMA, I don't see any timeouts, I don't see any interrupts.
I could write the same code in C and it would be just about as long.
