Am I the only one who feels Mozilla is losing focus? It's not that I think they shouldn't experiment (that's great for a non-for-profit company), but it seems they never really decided where they boundary ends. So from time to time you hear that they went a _little_ bit further from their core: it's always a little bit, but these sum up.

And I am not saying this to bash Mozilla: I think they are essential for a good Web. But that's exactly why I am concerned they are spreading themselves too thin.

WebGL is a nice compromise. But, coming from a console game development background, I won't be happy until both Mozilla and Safari adopt NativeClient. C++ with Posix and OpenGL can do amazing things on the iPhone. On even a netbook it could be 10 years ahead of the best HTML5 demos.

For example: WebGL can run Quake2, but NaCl could easily handle Epic's "Project Sword" http://www.youtube.com/watch?v=x2ncmwgJzqw

I think you're underestimating WebGL. The Quake 2 demo was a port from C to Java, compiled to Javascript, of a codebase that was originally designed in the heyday of software rendering. It's surprising that it works at all, and shouldn't be seen as a demonstration of the full capabilities of WebGL.

That "Project Sword" demo is built with OpenGL ES 2.0, which is the same API WebGL is based on. All the GPU-based effects will work exactly the same in WebGL. The only part WebGL might have trouble with is the CPU-driven particle systems and cloth simulation, but even there you might be surprised at what V8, Nitro, and JaegerMonkey can achieve these days. And if the performance still doesn't satisfy you, GPGPU is now an option.

NaCl is cool technology, but it won't be ready for wide adoption until it's divorced from x86.

From http://code.google.com/p/nativeclient/:

"Some ARM and x86-64 support is implemented in the source base, and we hope to make it available for application developers later this year."

The port to ARM and x86-64 are pretty much done by the way. Also, see:

http://code.google.com/p/nacl-llvm-branches/

Very true. Since it has been x86 specific from the start, that will be hard to do. PNaCl is trying to get there, by making LLVM architecture-agnostic, but it's an experiment at this point. And even if it does end up working some day, it may or may not get into other web browsers than Chrome, not to mention iPhones.

Personally I suspect JavaScript (or a subset of it) will get to the speed of native code sooner than we can run native code in a cross platform way in most of our browsers.

I know plenty of entries from the Node Knockout competition that fit this description perfectly. I wonder if Mozilla was inspired by what they saw at Node Knockout to put this on? There were a bunch of high-profile Mozilla employees judging.