someone interested in packaging VirtualBox?

[Les Mikesell]

Daniel P. Berrange wrote:

>>>>> It should use the KVM module driver, and if the current functionality in
>>>>>> KVM is not sufficient then VirtualBox should work with upstream to 
>>>>>> address
>>>>>> the limitations. Having multiple kernel modules for virtualization does
>>>>>> not help anyone.
>>>>>>
>>>>> Well, that would be quite a challenge.
>>>> Especially on x86 32-bit processors that KVM doesn't support... Or for 
>>>> people who want the option of moving their virtual machines to a windows 
>>>> host.  The functionality doesn't seem the same at all.
>>> I didn't say it was easy - just that if you ever want VirtualBox to be a
>>> part of the mainstream Fedora kernels it is going to have to stop 
>>> duplicating
>>> functionality already present & work with Linux kernel developers. What
>>> VirtualBox does for kernel drivers on Windows is utterly irrelevant & need
>>> not share any code with the Linux support, nor mandate what the Linux
>>> support looks like.
>> And what you are saying here is irrelevant to people who want their 
>> virtual machines to be portable.  KVM simply isn't useful to them and 
>> you make fedora less useful as well by not including virtualbox. 
> 
> That's a very short term view.

No, it's a user's view with no interest in being limited to single 
platforms or limited functionality.

> History has shown time & agin that betting
> on technology that is not in the mainline kernel brings severe long term
> maintainence pain which is not sustainable.

The pain of interface changes in the kernel is self-inflicted. And it is 
just one of the reasons users should stick to things that work 
cross-platform.

> Most spectacularly this is
> demonstrated by the Xen kernels. Including Xen in Fedora gave us some
> short term wins, but it has been a HUGE timesink diverting valuable
> resources from more useful development efforts. KVM would be alot further
> forward were it not for the resources spent keeping out-of-tree Xen kernels
> working. 

The Xen concept of the virtualized guest needing to know it is 
virtualized is flawed anyway.  You'd have to expect problems from that. 
    Handling it correctly on the host side should make it equivalent to 
the vmware module that doesn't need any specific management in terms of 
the rest of the kernel.

-- 
   Les Mikesell
    lesmikesell at gmail.com