To libvirt or not to libvirt?

[Technologov]

I'm developing GNS3 Network Simulator - we are a cross-platform and multi-hypervisor solution. (currently supporting 3 engines - Qemu/KVM, VirtualBox and Dynamips).
Considering to add OpenVZ and LXC hypervisors in some (distant) future.

One possibility I thought about was using libvirt instead of VBoxPython API directly. *Theoretically* it would allow us to consolidate Qemu/VBox/OpenVZ and LXC into one GNS3 back-end -- libvirt.
The problem -- doubt how well VBox is supported there.

And according to this post, I am quite right about my fears moving to libvirt:
http://www.redhat.com/archives/libvir-l ... 01080.html
VirtualBox 4.1 support in libvirt came only in october (!), 3 months after v4.1.0 was released (July). And even this was not tested.
By contrast, GNS3 added support for VirtualBox 4.1 in May, two months before it was released.

Any opinions ? User experiences of libvirt + vbox combo ?

History:
Original vbox driver for libvirt was done by "Pritesh Kothari" from Sun. No idea what was his goals, but now there is no active maintainer.
libvirt v0.9.7 clams to support VBox 2.2 / 3.0 / 3.1 / 3.2 / 4.0 / 4.1. Looks good on paper, but doubt if it is realistic to test them all. More realistic goal would be supporting last 2 releases. (4.0 and 4.1)

libvirt IRC channel + archive:
#virt on irc.oftc.net
http://www.redhat.com/archives/libvir-list/index.html

-Technologov