Open Stack

In response to https://review.openstack.org/#/c/12413/ I'm faced with the following question: should openstack common modules be included in cases where they're not providing any benefit other than standardization?

In particular, there are 3 points around that review:

  1. The jsonutils code is designed to solve problems Horizon doesn't have.
  2. The timeutils code duplicates functionality already built into Django (which Horizon is built on) and is only included because jsonutils now requires it.
  3. The timeutils code is now adding external dependencies (iso8601 python module) which *again* is duplicating built-in function in Django.

This seems like a really bad approach. It's a long dependency chain that adds nothing other than saying "we're using openstack-common".

I'm inclined to remove jsonutils from Horizon (the "setup" module from openstack-common is still quite useful) and just go back to using the built-in python json.

Alternatively, we need to introduce some constraints around introducing dependency chains in common...

Thoughts?

    - Gabriel