Are Linus Torvalds, IBM, AMD, others at odds over pushing hardware drivers out of Linux?

I can't possibly profess to know what the pros and cons are of pushing hardware device drivers into the hypervisors that support virtualization technologies like those from XEN and VMware. But it's clear from a recent thread over at the Linux Foundation that there's some disagreement amongst experts (including Linus Torvalds himself) as to whether it makes sense or not. Here are some highlights and excerpts from the discussion thread so far:

Marc J Miller (AMD's Open Source Relations Manager): ....If driver management becomes the role of the hypervisor or even of the firmware and there is no "host" operating system to speak of, will Linux drivers still matter?.....

Gerrit Huizenga (Architect at IBM's Linux Technology Center): ...In short, pushing the problem down to the hypervisor will help in some cases, but will probably make things harder in a few other cases. Instead of pushing drivers to mainline Linux and the distros, we'll all have to push them to the latest hypervisor (Xen, Linux/KVM, VMware ESX, etc.) before the software stack can take advantage of the latest hardware.....

Dave Jones (Codemonkey): Contrary to popular opinion, not everyone wants or needs virtualisation. The performance hit of that extra layer of abstraction well never completely go away regardless of how many shiny new features each generation of CPU gains....

Why would I possibly want to use up a bunch of memory, and CPU for functionality I have absolutely no use for, whilst adding another potential source of bugs?

Linus Torvalds: I think what you're seeing is virtualization proponents being absolutely _desperate_ for any reason to use virtualization....The fact is, the absolute last place you want to see drivers is in the hypervisor....So next time you hear about hypervisors doing anything at all, ask yourself where the message comes from. Does it come from a group of people who are desperately trying to make themselves appear relevant? Or does it come from forty years of actual real-life practice?....I don't doubt at all that virtualization is useful in some areas. What I doubt rather strongly is that it will ever have the kind of impact that the people involved in virtualization want it to have.

Whoa! Strong words from Linus regarding the virtualization community!

Far be me the one to summarize here. The issues seem a bit beyond my technical capabilities (OK, a lot!). But if device drivers can successfully move into virtualization hypervisors, does it not call into question the relevance of operating systems overall (as opposed to the relevance of hypervisors)? And regarding the supposed "irrelevance" of hypervisors, tell that to Jeff Bezos and the crew at Amazon who run that company's Elastic Computing Cloud (built on XEN virtualization, as I understand it).

In some ways, this harkens back to my question regarding how much desktop operating systems will matter by 2010 (when the next version of Windows desktop is expected).

When I asked that question, I simply assumed that the role of operating systems in interfacing to the hardware was becoming commoditized while applications moving into the network could be diminishing the value of the rest of the APIs and services found in an OS. Even then, the OS still has a role. But if the role of the app services and APIs found in most operating systems starts to give way to network bound applications and the hardware drivers begin to move into a layer below the OS (eg: a hypervisor), what's left for the OS? Hmmmm.

Or maybe I'm barking up a tree that I have no business barking up? You tell me.