A few weeks ago some interested parties had an informal meetup at the Mozilla office to discuss the issues surrounding the Gamepad API as-implemented. The biggest issue we noted is the inconsistent mapping of controls across browsers and operating systems. We talked at length about this, and it feels like a really hard problem to solve. A number of people have written content libraries to try to smooth this over (including myself and Scott):
https://github.com/luser/gamepadmapping.js
https://github.com/sgraham/gamepad.js
http://mozilla.github.io/galaxy.js/docs/galaxy.js.html#getGamepads

However, because of the differences across browsers and OSes the complexity blows up quickly. (Example: https://github.com/luser/gamepadmapping.js/blob/6dae0a5751df4acacfae2a5d3619758b7bb4bf43/gamepadmapping.js#L90 )

Specing things to remove this complexity might be possible, but it would be hard. We think as a stopgap we can make things better by simply exposing more information about the controller and the source of the data in a structured manner, to allow content libraries to make better choices.

Right now there's an "id" string on the Gamepad object that is basically unspecified (bug 25202). We should deprecate that in favor of a dictionary of useful attributes. My proposal would be something like this:

dictionary GamepadAttributes {
  unsigned short? vendorId;
  unsigned short? productId;
  DOMString name;
  GamepadDataSource dataSource;
};

enum GamepadDataSource {
  "hid",
  "xinput",
  "linuxjoy",
  // other values
};

interface Gamepad {
  GamepadAttributes attributes; // Probably needs a better name
  ...
};

We'd like to see how well we can do with content libraries if we give them better data, and this feels like it ought to be enough to work with.