openstack-poc team mailing list archive

[Ewan Mellor <Ewan.Mellor@xxxxxxxxxxxxx>]

Firstly, apologies for not making the PPB meeting today.  I was actually locked in a meeting about OpenStack API support, which is ironic given the topic of conversation at the PPB today.

There is one thing that we haven't discussed recently, which I want to raise.  I don't think we've made any statement about API stability beyond Diablo - i.e. support for Diablo-based clients running against Essex and beyond.  I am hearing that people have bashed their head against the wall so many times with Cactus -> Diablo transitions and intra-version transitions on Diablo that they have given up, and won't try again until there are large (i.e. Rackspace Cloud-scale) production deployments of OpenStack to guarantee the stability of the API.  This is not good for any of us.

I have come to the conclusion that, for the good of the project, we have to publicly guarantee that clients written against Diablo will work against all future GA versions of OpenStack, and I think that the PPB would be right to issue a diktat to that effect.

I don't care about Cactus, because it was never production quality, and I don't care so much about Glance because it's a smaller API, but it's essential that clients can start to rely on the stability of the Compute and Object Storage APIs (including being able to log in, which means Keystone has to be stable too).

Obviously the API can continue to evolve - it just needs to do so in a backwards-compatible (or versioned) manner.

Also, extensions would be excluded from this - caveat extensor, or something.

Thoughts?

Ewan.