These are major and invasive changes to POSIX. No reasonable person would expect to be able to do things like change PID semantics or shared memory. Yes, it might solve the problem that they sought to solve. But I would be very surprised to see this meet with any large-scale deployment. It's better to work with the system than to just arbitrarily decide Unix is wrong and rewrite it.

When using Xinerama (which you really ought to be if you want control over your multi-screen setup), many tiling window managers can do all sorts of neat things. I personally use Awesome, although I'm told that xmonad is also good at this.

Buh? As far as I can tell, Google doesn't have a platform called "AppLogic". Perhaps they were referring to App Engine? And it's not even the editors' fault this time -- TFA has the terms wrong too. That really inspires confidence...

I've found that Verilog is much easier to learn and teach (at least, for an undergraduate engineering-type class). But, as others have mentioned, do NOT think of it as a programming language. Think of it as a convenient way to draw schematics, a very sophisticated keyboard layout for Xilinx. You should ALWAYS write synthesizable code for everything except your testbench, and you shouldn't have to synthesize it to know what it looks like. As to why not VHDL, well, VHDL is the COBOL of HDLs. Way too verbose for me.

Yeah, that seemed a little unlikely to me as well. There are 2**128 addresses in IPv6. Even assuming that all of these were allocated in 64-bit subnets (fairly common), that's still 5*10**15 subnets. Which is a hugely ridiculous amount, many times larger than the IPv4 Internet. Something's fishy about this number...