openstack team mailing list archive

[Brian Waldon <bcwaldon@xxxxxxxxx>]

tl;dr - We're going to use minor versioning for the v2 Images API and take a more iterative approach to API design and implementation

Up until this point we have depended on big design up front for the Images API. We spent most of Essex talking about v2 without writing any code. At the beginning of the Folsom release cycle I took it upon myself to start implementing the v2 Images API (which has been rather slow-going). Several people have helped, for which I am grateful, but ultimately we didn't finish what was needed. Not only did we not finish everything, but there are several things in the spec that need to revisit.

The (possibly incorrect) assumption I have been working under is that we would release the v2 API and that would be it - just a major version, no minor versioning, no extensions. If we wanted to make changes, we would put them in v3. 

That won't work. It has taken two releases to get to this point and we still aren't done. The solution as I see it is to use a more iterative design/implementation cycle for our APIs. We can't be designing massive API specs up front - it doesn't always work. We need to have the flexibility to move fast and get to something that works that we are proud of. 

We are going to use minor versioning for the v2 Images API. This means that the Folsom release will implement a v2.0 spec, with additional (and already identified) features in a v2.1 sometime early in Grizzly.

Please voice any and all concerns ASAP since we're on a pretty short timeline.


Brian Waldon
The People's PTL