Let's just say that comparing Windows to the linux kernel would be a mistake. Hosting the windows source would be a challenge for any source control system out there. Git / github is not exactly optimized for projects of that size.
But wouldn't it be nice if Windows could be modularized in the process of such an open-sourcing effort, such that that comparison wouldn't be a mistake? For example, if just the NT kernel could be separated into its own project, with a clean linkage-point, there'd probably be all sorts of interesting efforts to, for example, create a Debian/NT distribution, or cross-pollinate driver code between Windows and Linux. (Actually, having two major open-source consumer-PC OSes would probably finally force a standardization on a HAL to allow a single driver-pool code base to be shared between the two systems.)
Sure, Windows is more than a kernel. I imagine the NT kernel vs the Linux kernel would be roughly comparable. But to compare 'Linux' the OS to Windows would mean Windows vs. GNU/Linux/X.org/Gnome/KDE... or say, the source for every package in the base Debian distribution.
Exactly. Now imagine that every part of it was just as bloated and convoluted as you'd imagine.
I doubt that Windows is monolithic/bloated. Convoluted sure, backwards compatibility is a bitch, but Windows is probably broken into neat little modules by now.
Apparently you don't know what you're talking about. Source Depot at Microsoft (rebranded Perforce) can do it, while TFS cannot, therefore half of Microsoft stores its source code in TFS and half in SD. Git scales really well, at least well enough for the Linux kernel which is not much less than Windows kernel.
What Perforce is good at and what Git is good at don't overlap too much.