I would add specifically that virtualization is overdue. Apple's strategy back in the day was pretty good: sandbox the entire old OS so that users can still get most things to "just work", but then support only a subset of APIs on the new OS. Apple also kept the old OS as-is (i.e. ugly), while using all the nice new technologies and look-and-feel as bait to encourage developers to move.

Yes very good point. Build a good foundation OS and then virtualise for backwards compatibility.

I think .net is a good bait for such a new system. From the little interaction I've had with C# devs, they seem to like it.

There is already the XP mode in Windows 7, which boots a virtualized XP for compatibility.

Thanks. Didn't know that.

Hopefuly this means that implementing the idea I'm talking about is not too much of a stretch from the current code.

What would you consider a maximum acceptable boot time?

As low as possible, and ideally virtually instantaneous. From a cold boot, I'd put a cap of 10-15 seconds tops.

Why would anyone reboot a system, instead of putting it to sleep¹? (Except when restart is required by an upgrade.)

[1] Powered-off-and-ram-image-stored-to-disk type of sleep.

My work laptop (Vista `Enterprise') takes longer to wake from hibernate than it does to cold boot.

But that's hibernate - Sleep takes less than 2 seconds to recover from for me.

I don't know how Apple does it, but both my iMac and MacBook wake up in a second. I guess they keep the ram powered up and only save it to disk (hibernate) when running out of battery (or after certain amount of hours?). Anyway, they are totally silent and consume very little energy when sleeping. That's how it should be.

I love how the difference between sleep and hibernate is abstracted away in Apple's implementation.

Yep, Apple laptops keep the RAM powered up until the battery level dips down to 20%, then it writes the contents to disk when the machine sleeps. It's called Safesleep and you can tweak this if you'd like by using a utility called Smartsleep.

Actually, according to the following sources, the content is written to disk _before_ entering sleep. This can be done very quickly because most of the data is already in swap file.

http://support.apple.com/kb/HT1757?viewlocale=en_US http://www.macworld.com/article/53460/2006/10/safesleep.html http://moishelettvin.blogspot.com/2006/11/windows-shutdown-c...

The OP specifically mentioned saving RAM to disk. I don't trust the battery in this work laptop to hold the state in RAM overnight.