Limitation is in "developers", that is WEB developers playing with embedded systems. There is nothing stopping you from deploying dedicated OpenGL application on RPi, but this is not the 2.0 way. Bros want nodejs and responsive UIs in javascript .... you end up with this: https://youtu.be/tG70c3Zadek?t=10m49s

tldr: full minute to boot X and chromium just to implement simple 20 lines of python application.

Don't be a dick about it.

Of course you can build an OpenGL ES application on the Pi – there are people doing this. But that's a absolute frigging world away from being able to build a usable UI on it.

For example, I am building a device that interacts with a user. It basically looks like a standard iOS or Android app, in that it has a few screens of text, some form fields, select boxes, and other UI widgets. Building this sort of application as a barebones OpenGL implementation would be stupid.

Whatever floats your boat. If you can leverage existing code then this might be an excellent option. But for the general use case building a simple and/or decent user interface I would say that you probably would be way better of just using QT.

Use MOAI. I did. It freakin' rocks on the rPi and is one hell of a great way to build a delightful UI ..

The most important thing is shipping ideas. Languages and system cannot be obstacle

Using the web with embed system allow new horizons and new people play with this. Of course, we can do applications with OpenGL but the first step can be with a universal environment like a web.