On Fri, Jan 11, 2013 at 10:05 AM, Noa Resare <n...@spotify.com> wrote:
> (This email is probably mostly for Wido and possibly Hugo but I'm sending
> it to the list to get a bit more transparency)
>
> Looking at the packaging branch(1) I think the work to reduce the number of
> subpackages is a very good initiative. The current setup for 4.0 with lots
> of packages with somewhat counterintuitive naming adds complexity both to
> the build process and to users of the package.
>
> The current state of the packaging branch contains subpackages -server,
> -scripts, -setup, -python, -agent, -system-iso and -usage.
>
> I think that packages should be as few as possible but that we should avoid
> payload duplication. In practice that would translate to one package per
> server use case (in my case, agent and server) and have shared artifacts in
> a common package.
>
> If we are doing any change where subpackages will be obsoleted we would
> need to follow the http://wiki.debian.org/Renaming_a_Package guidelines
> (which means creating an empty last version of the old package depending in
> the new package to get the upgrade path properly) and if we are doing that
> for any package we might as well do it for everything and bite bullet with
> the cloud -> cloudstack renaming.
>
> So, my proposal: three new packages:
>
> - cloudstack-server (containing everything specific to the server)
> - cloudstack-agent (containing everything specific to the agent)
> - cloudstack-common (containing the things needed by both)
>
> and a separate source package emitting new empty versions of all current
> packages that could be added once to the repo referenced on
> http://incubator.apache.org/cloudstack/downloads.html
>
> /noa
>
>
> 1) https://github.com/noaresare/incubator-cloudstack/commits/packaging
>


So while I like the idea of consolidation, in many ways I think some
of the javelin refactor work may mean we end up needing new
subpackages to deal with the new modular pieces of things that would
once have been in -client -client-ui -server, etc.
Alex: any insight you can give to how you see the various pieces being
deployed once javelin is in place?

--David

Reply via email to