If you read through the depths of Magic Leap's site combined with their patent applications, it becomes clear that they are trying to develop a set of goggles which combines some form of projection onto the retina [0] with some form of selective blocking [1] (to give contrast, and prevent the projected images from appearing as hazy mirages over the light otherwise reaching the retina).
Especially the blocking would be impossible to achieve with anything but goggles.
[0]: https://www.google.com/patents/US20140071539 [1]: https://www.google.com/patents/US20130128230
Still unclear. They do seem to imply they want it to be an unconstrained mobile AR device, but that is indeed ambitious enough to warrant skepticism. Walk-around tracking for home/anywhere is still an unsolved problem for Oculus, and overlay AR is at least several times more demanding.
That's actually how (kinda) Stephenson's VR system he describes in Snow Crash works.
This kind of gets away from many of the issues you have with the Oculus Rift where you have such a tiny window of time (20ms or so) to react to how the person is moving his head. You still have to change the image based on the user's position, but not as much on head rotation which is the really hard part. Just each eye's location in space matters.
Multiple users would likely require multiple projectors.
But I mention the "they're trying to build like the equivalent of the early table-based Microsoft Surface (before that meant a tablet), but pseudo-holographic instead" idea just because it's so unbelievable to me that stand-alone goggles can possibly deliver what they're claiming.