[kidmenot]

> This is what I hope should happen with Windows. They should decide on POSIX kernel and completely rewrite most of it.

They simply can't afford it, not because they lack the resources but from a strategic point of view. See the other comments talking about compatibility. Windows works for many, many, many people and businesses. It would be a suicide move. People, and even more importantly businesses, don't jump on any hot new thing just because it's available, because change has huge costs in an enterprise context. If Microsoft ever tried to pull the rug from under them, you can be pretty sure that would be the end for them.

[norea-armozel]

I agree that they're not likely to do a full abandonment of the Windows stack, it still makes money for them so why mess with a good thing? I think more likely is that as Azure expands they'll try to make more money off of that platform and use Windows (desktop/server) as their supplementary income which in itself is a big gamble IMO.

[overgryphon]

Azure runs on Windows.

[newjersey]

> If Microsoft ever tried to pull the rug from under them, you can be pretty sure that would be the end for them.

It is necessary though. Microsoft seems to be preparing for a post-Windows eventuality as well. Doesn't mean they have to write their own operating system.

Edit: it is mostly for azure anyway... I think Microsoft might as well ride Windows as long as it can and give it up.

[faragon]

This one's pretty obvious if you've ever written a malloc. Compare VirtualAlloc² with the mess of mmap/madvise/mprotect. I'm not even sure you can get the nice explicit reserve/commit of Windows with pure POSIX APIs.

Then, can you explain why the C library provided by Microsoft in their compilers has such slow 'realloc()' implementation? (e.g. expanding a 10MB allocation to 20MB using realloc())