The guy who invented the Oculus started out by giving simple instructions to build your own:
http://projects.ict.usc.edu/mxr/open-source/fov2go/

A smart phone and some 35mm slide 3D glasses gives you a very compelling experience.

This really is superb. Has a really nice Android client as well as the web interface.

A good video card is connected to your CPU via a 32GB/s bus. Either you have a very good network, are wasting your graphics card, or you aren't going to get anything close to network transparency.

Costs aren't only the license fees paid (and this is operating outside any copyright territory so licences fees would be extremely prohibitive to enforce anyway).

The biggest cost of Windows/OS:X is that can't make changes. A satellite may well have hard real-time requirements or require other kernel changes that exist for Linux but not for closed source general purpose OSs.

Against who? Anyone who can field anything that could even shoot in the general direction of a single F-22 also have nukes.

Passwords should (and usually are) stored as hashes which means you can very quickly hash the user's entry and compare if it is exactly the same as the password, but by design can't infer any other details about the password if the entry is wrong.

Anything that allows you to compare how 'close' an entry is to the users current password is obviously makes guessing that password far easier.

If your passwords are securely salted and hashed then storing additional old entries shouldn't lower security, and as you say ensures that the user can't reuse a password precisely, but any minor change to the password with result in a completely unrelated hash.

Anyone who can watch 5 minutes of Top Gear without realising it is a sitcom couldn't afford a Tesla anyway.

The is no requirement that the author pays, only that someone dose. Authors who can't afford $99 only have to find someone who believes their paper is worth at least that.

So I can just make my own TPM chips based on the standard right?

Way to set really low goals.

How would micron-LEDs allow more channels? If you are using different frequencies of light you wouldn't need them that small to cover the usable spectrum. If you are planning on resolving each pixel form the receiving device you certainly couldn't use micron sizes over any useful distance?

Why dose migrating to another server disconnect people?
Surely in a case like this they can just pause the action (if they can slow down time surly pause is easy) and tell everyone it will resume in 10min as the do the migration.
Worst case you could disconnect them but tell them if they don't connect to the new server within 10min then they are treated the say way as if they had yanked their broadband cable?

The goal of Minecraft et al. is to provide very large (or sometimes truly infinite worlds). To achieve this you have to use extremely efficient memory storage. I don't know the actual figures but I assume Minecraft stores one or two bytes per voxel, which are approximately 1 cubit-foot, so a very efficient representation.
This simply isn't enough memory to store more then a few shapes (though a sphere dose seem a good choice).

But Micro-USB cabled still aren't anywhere near as reliable, we need to switch back.

Unfortunately that is an array of strings not numbers. However more importantly you have also made the cardinal sin of making assumptions about the specification. The fact that they defined the array had to contain 1,2,3,4,5 dose not imply that it had to contain them in that order.