An un-Folsom morning

[Lars Kellogg-Stedman]

On Wed, Nov 21, 2012 at 03:59:53PM +0100, Alan Pevec wrote:
> I'd recommend you to subscribe.
> OpenStack has not declared any long-term-stable release upstream yet,
> so we'll continue doing  updates in EPEL until such release exists.

That seems like a good indication that OpenStack should probably not
be in EPEL proper.

We have always believe that the "Enterprise" part of EPEL meant that
updates would not cause this sort of disruption in a functioning
environment.  Frankly, even if we had seen the announcement our only
solution would have been to exclude all of the openstack packages from
updates; we're simply not in a position to invest the time required to
make the jump to Folsom at this time.

If we wanted to be running a bleeding-edge environment with disruptive
upgrades we'd be running Fedora instead of a RHEL-derived
distribution.

For something moving as rapidly as OpenStack it seems best to either:

(a) Name the packages after the release.  Have openstack-essex-nova,
openstack-essex-utils, etc., and then introduce openstack-folsom-nova
and so forth.

(b) Remove the packages from EPEL proper and place them into
release-specific repositories.  Someone running the Essex release can
track the epel-openstack-essex repository, someone who doesn't mind
their entire production environment exploding on a Tuesday morning can
track epel-openstack-current.

Either solution would eliminate the sort of problems introduced by the
Essex->Folsom upgrade in EPEL, and (a) would have the advantage that
it wouldn't require additional repositories.

-- 
Lars Kellogg-Stedman <lars at seas.harvard.edu>  |
Senior Technologist                           | http://ac.seas.harvard.edu/
Academic Computing                            | http://code.seas.harvard.edu/
Harvard School of Engineering                 |
  and Applied Sciences                        |