Code review comment for lp:~ev/ubuntu-ci-services-itself/restish-swift-deployment

On Tue, Jan 28, 2014 at 04:54:17PM -0000, Andy Doan wrote:
> On 01/28/2014 10:48 AM, Evan Dandrea wrote:
> > .bzrignore? :)
>
> yeah - i thought about that. its not a terrible idea. i suppose that's
> not any worse than my hack to copy the charms to a temp directory

I'm not a huge fan of the .bzrignore approach. I agree with Andy that
having external charms in a working copy of a bzr branch is bad. I like
to be able to easily blow away the juju-deployer set up charms/precise
directory at will and not have to pick and choose what to delete.

I don't have an easy solution either. I think having our charms in a
separate branch might be a better approach especially if we are also
going to need to be able to grab a specific revno of a charm. For that
matter I even think one branch per charm might be the best way to go.

I think the reason we initially went with including the charms in the
ubuntu-ci-services-itself (u-c-s-i) branch was to avoid having to update two
branches in lock-step. If I'm mistaken the rest of this paragraph is
probably moot. What solutions are available for inter-charm
dependencies? i.e. charm-A revno X requires charm-B revno Y. If we're
using revnos for each charm in u-c-s-i's juju-deploy config won't this
mostly mitigate the issue?