Lots of new layers!

[Stuart Bishop]

On 6 January 2016 at 22:48, Marco Ceppi <marco.ceppi at canonical.com> wrote:

> We now have 20 interfaces and 18 layers published to the index:
> http://interfaces.juju.solutions - this is amazing, and I know there are
> even more layers out there in flight still. With the sudden burst of layers
> (both charm and interface) we've started seeing charms in the review queue
> built on layers. Cory sent an email out earlier[0] about how to handle the
> submission of charms built with layers, which is great.
>
> I wanted to start discussing how we should handle the curation of these
> layers. Ideally I'd like to follow something along the lines of the charm
> review process, where charms are reviewed to make sure they promote the best
> possible practice. This helps to make sure deployments are high quality and
> means any charm could be used as an example of how to write a charm.
>
> Applying this to layers seems like a good start, but I don't want to
> overburden the current review process and I don't want to stifle the
> positive trajectory of layers being created. One idea, which I think would
> be simple enough is to continue to keep an open registry but simply have new
> layers announced on the list by the authors. This will help expose the
> layers being created, invite feedback from the community to the authors, and
> give us a chance to make sure layers implement the latest workflow since the
> charm build process is still evolving.
>
> Opinions?

I think you need to keep curation separate from code reviews.
~charmers does not have the bandwidth nor domain knowledge to handle
charm code reviews as it is, and splitting things into layers has
already accelerated development and will make this problem worse. So
curate away, organizing and overseeing the effort, but don't put up
roadblocks to the development process. Instead, make your decisions as
curators based on the quality of results produced by the layer and
charm developers (who can define their own review and landing
processes), backed up by community opinions and automated test
results.

Now, having got the reactive rewrite of the PostgreSQL charm ready for
review, I now need to split out the three layers that will be shared
with the Cassandra charm and available for general use... all this new
stuff seems to be working :)


-- 
Stuart Bishop <stuart.bishop at canonical.com>