Megatexture is not so much to do with large textures or open areas, but more to do with virtualisation of video memory. This virtualisation does allow these things for sure, but it also allows one item you failed to mention, which is for texturing to be unique. This means no more tiling, no more repeated textures. It's just a hunch of mine but I strongly suspect that it's also going to be
very performance-efficient. Carmack himself has indicated that the technology behind it would have worked on a Voodoo 1, and common-sense tells me that reductions in state changes and opportunities for greater draw call batching would result in something that - for the same polycounts - runs considerably faster than anything else out there. Of course that will be offset by higher polycounts and more complex visual FX elsewhere.
What you're calling a "megatexture" in your Google result looks a lot more like a texture atlas to me. That's a very different thing, so I'd caution you to lay off the negative comments until you have your facts right.
Quake-level technology actually
can do wide open areas quite well. It takes a bit of coaxing in engine code, but it is possible. Scene traversal is a major CPU-side bottleneck, and the quite heavy tesselation that comes from QBSP doesn't help things much. These things do have solutions, and while the end result is not as good as alternative approaches, it's nowhere near as bad as it's commonly assumed to be.
You shouldn't see skeletal animation as being some kind of holy grail of animation methods. It's just a different set of tradeoffs, and for some classes of model frame/vertex animation is still very much the right way to go.
Regarding idTech 6, there is certainly
talk of voxels at the present moment, but it's far too early to be saying things like "idTech 6 uses voxels". To quote from the man himself (written in early 98):
I'm not ready to talk specifically about what I am working on..... Quake went through many false starts (beam trees, portals, etc) before settling down on its final architecture, so I know that the odds are good that what I am doing now won't actually be used in the final product, and I don't want to mention anything that could be taken as an implied "promise" by some people.
Only that one is done in much bigger scale and there is some method to uplad texture data from it to graphic card (but You are uploading parts of this texture atlas into graphic card too; but sometimes whole atlas is uploaded). I know it is bigger than graphic card can have, but to me idea is the same, have lots of textures packed into one as You can see on pages 9-12 (some of them being huge and bigger than graphic card can handle). AFAIU Carmack just made this texture atlas in big.
i can allready say with 99.9999% certainty that about 95% of the engine memory operations are from loading textures 