Let's do it now with this release.

Dual-maintaince of separate ways to do it leads to certain paths not being well tested, at it sounds like we already have
that problem in some corners.

Folks with issues about what virt-install can't do can then take it up on the virt-install lists, and cobbler can just
be (for virtualization) the adapter layer that instructs virt-install what to do (obviously keeping the reinstall and config
stuff it does as well).

If bugs arise, we can fix them.

Make it so.

On Monday, February 6, 2012 at 3:51 PM, Cole Robinson wrote:

On 02/06/2012 03:46 PM, Michael DeHaan wrote:
Hmm.

Another blasphemous thought might be to just shell out to the qemu and Xen
command line tools directly if preserving these options are important.

Someone would have to do the legwork of making it though :)

Well, like I said before, I'd be okay with changing the patch to preserve the
original code if 'import virtinst' still works, and my new code would be how
koan provisions libvirt VMs on fedora17 and newer. Presumably people on RHEL5
are updating cobbler but not virtinst, so they will be using the same old code
paths.

This would also mean I have to worry less about regressing everyone's systems,
and only regressing for people that are also upgrading their host to another
major OS (say f16->f17)

However I would then recommend yall come up with a timetable of when to drop
the old virtinst code for good, and make the virt-install impl the canonical
way to provision libvirt guests, just to save yourself headaches. That would
give you time to get out ample warnings, announce it loudly, etc.

- Cole