First I want to congratulate the Kronos group for the OpenML initiative. I find it very credible, and I am sure this is a major step for the evolution of digital media and GUIs. I would also say : it was about time !!! We all needed some kind of cross platform directX, this has always been a major issue for the industry.
But OpenML makes, to my opinion, a huge omission : there is no proposition for standardisation of user input devices, such as joystick, mouse, and keyboard with low level access.
From the game programer's point of view, this is a major issue, because, whenever we want to port a game, this is code we have to rewrite, and it is different for each platform. This is like reinventing the wheel every time.
Not only talking about games, there are many applications requiring the use of joysticks while designing sound for films : operators make the sound move in 3D for dolby 5.1 using joysticks.
Think about it : With user IO discovery, you can program 100% of a game using OpenGL and OpenML : user IO, graphics and sound.
Without user IO discovery, you can forget about the OpenML portability, and have to go for platform specific code.
After all, why wouldn't jostick's makers release OpenML drivers, so joysticks could also be used on Linux and MacOs (which would open them new markets), just like 3D boards manufacturers release OpenGL Drivers ?
At last, unifying APIs for Network and 3DSound would also definatly be a big plus to OpenML...