Hasn't software mixing been in ALSA for ages, with the dmix plugin? I don't think Pulseaudio is necessary for everything.
I do see the Networkmanager is a necessary evil, but why does it have to be so awful? It's great if it works, but if it doesn't then solving the problem is extremely hard because of the bloated, uncommunicative nature of the program.
Pulseaudio can abstract over multiple devices, allowing you to mute individual applications, move streams between physical pieces of hardware, and so on. Audio mixing in kernel-space is brittle and the wrong place for the code anyway. Doing it in userspace is more secure, more flexible, and generally a good design.
In theory, maybe. In practice, Pulseaudio is just horrible and gave me troubles I have never experienced with ALSA. I'm personally convinced it's one of the biggest mistakes made in the Linux world.
However, if we're going to put the sound server in userspace, which isn't a bad idea, the kernel side should be broken down. This would be an excellent time to deprecate ALSA!
I do see the Networkmanager is a necessary evil, but why does it have to be so awful? It's great if it works, but if it doesn't then solving the problem is extremely hard because of the bloated, uncommunicative nature of the program.