You're starting from the assumption that they're mostly concerned about some devotion to standards and interoperability and that they wouldn't just make their own proprietary alternative. I think it's far more likely that they'll move in the direction of a partitionable fully managed environment running on top of a very thin hardware abstraction layer. It'd be the equivalent of booting the computer into a .NET or Java VM, but without anything in between that and the hardware other than an exokernel. The Midori project is doing something like this, and speculation points to Azure extending that model to provide utility/cloud computing.

Microsoft would be able to make a clean break from the older APIs, leaving Windows 7 as the legacy OS able to run .NET apps on top. Azure would be the OS which runs .NET apps natively and could pool dozens of heterogeneous commodity servers into one virtual megaserver. Linux would have Mono, but there's nothing that would compete against it directly until someone could formulate an alternative. In the meantime Microsoft would have a fairly lucrative market to themselves while drawing more people to .NET.

Windows hasn't used that security model in almost a decade now. Now it uses a role-based security model which provides the same protections as the Unix group-based model, albeit more flexibly and with finer-grained control. That's proven itself to be quite robust: The problems arise largely from code execution exploits in software using the network like Internet Explorer and from the users themselves. UAC and not running by default in Administrator mode help prevent some of the nastiest, but there's still a lot of nasty stuff that's still possible. Deleting a person's home directory usually hits them far worse than anything else because if they didn't back it up, it's gone, while the other stuff is just a reinstall away. Botnets only require being able to run software which can poll somewhere else to pick up orders, to send mail in some manner, to stick that program in a hidden directory with the same name as an important long-running system process and make it executable and finally to set it to run every time they log in. For a single user system run by the proverbial grandmother, that's just as effective as whatever crazy elaborate scheme any hacker can cook up.

COM objects. More specifically ActiveX. Or was that a rhetorical question?

The RTC is used for a variety of purposes even if it doesn't need the current year. Anything where the player waits even for a few seconds uses that clock or the interrupt timer. Linux uses the interrupt timer by default in those cases, but it initializes the timer and synchronizes it against the RTC if it exists to protect against drift and to account for leap seconds and the like.

More pointedly, it's often our own ignorance which leads us to think the world is populated with idiots.

Tablets didn't fail so much as tablet PCs failed. Stuff like cell phones, GPS units, book readers, personal media players, PDAs, smart remote controls and the like are all quite successful and available as small tablets, but they're not PCs. Designing a bigger and better one of those is more natural than trying to make a laptop usable without a keyboard.

After 1.6_10 and JavaFX, they're effectively one and the same. There's a JavaFX demonstration of an experimental feature where you can drag the Applet from its spot in the page and it becomes a regular JWS application with only the slightest pause as it spawns a new OS-level container. I'm pleasantly surprised to see all that Sun's done to improve Applets even after they've mostly failed in the market.

I wait with bated breath for a hyperlink that I can click on to play an Ajax version of Quake 2. Until then I'll just have to make do with http://www.bytonic.de/downloads/jake2_jogl11.jnlp instead.