Open Stack

On Sun, Nov 24, 2013 at 11:10:23AM +1300, Robert Collins wrote:
> On 23 November 2013 05:32, Daniel P. Berrange <berrange at redhat.com> wrote:
> > On Fri, Nov 22, 2013 at 11:24:18AM -0500, Russell Bryant wrote:
> >> A good example is the current discussion around a new scheduling
> >> service.  There have been lots of big ideas around this.  Robert Collins
> >> just started a thread about a proposal to start this project but with a
> >> very strict scope of being able to replace nova-scheduler, and *nothing*
> >> more until that's completely done.  I like that approach quite a bit.
> >
> > I'd suggest something even stronger. If we want to split out code into
> > a new project, we should always follow the approach used for cinder.
> > ie the existing fully functional code should be pulled out as is, and
> > work then progress from there. That ensures we'd always have feature
> > parity from the very start. Yes, you might have to then do a large
> > amount of refactoring to get to where you want to be, but IMHO that's
> > preferrable to starting something from scratch and forgetting to cover
> > existing use cases.
> 
> That is precisely what I'm suggesting. Forklift the code to another
> place and put in place in nova enough glue to talk to it. Complete
> parity, no surprises on performance or anything else. Then start
> evolving.
> 
> Anything else is just risk for no benefit.

Great, that sounds like the exact right way todo things.

Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|