[openstack-tc] Policy on "3rd party" APIs and Nova

Mark McLoughlin markmc at redhat.com
Tue Nov 6 16:05:30 UTC 2012


On Tue, 2012-11-06 at 14:40 +0100, Thierry Carrez wrote:
> Anne Gentle wrote:
> > Regarding the scope clarification - my question is, is the TC making
> > policy that all -core projects must follow?
> 
> We currently have a PPB-old edict that prevents projects from accepting
> new 3rd-party APIs. Vish proposes to dissolve that edict and let all
> projects pick 3rd-party APIs if they think they make sense for their
> project, can be maintained etc.. So in a way, we would be removing
> established restrictions rather than making a choice.

It may be a pedantic distinction, but what I'm proposing (and AFAICT
Vish agrees) is just that we recognize what the PPB previously voted on
was aspirational rather than an edict.

i.e. the project still aspires to exposing "stable, complete, performant
interfaces so that 3rd party APIs can be implemented in a complete and
performant manner" and that "other APIs will be implemented external to
core" but that it will take time to get there.

I actually think that's all pretty self-evident.

> The reality is a bit more complex though. As Jay mentioned, the edict
> had for example the side benefit of making it easy to refuse 3rd-party
> APIs (no need to pick losers and winners). The edict was also about
> ensuring the OpenStack API was the first-class API, by encouraging
> projects to come up with a clean internal compute API for 3rd-party APIs
> to connect to.
> 
> While I'm all for letting more freedom to each project on that specific
> matter, I would really preferred if we arrived one day to that clean
> internal API... and I feel like removing the incentive to do it is not a
> step in the right direction.

Right, one can think of a "no more 3rd party APIs" edict as a way of
forcing individual projects into creating the "stable, complete,
performant" interfaces necessary.

What we're seeing in Nova is that it is not proving to be a positive
incentive for anything. It is merely frustrating people's (most
excellent, IMHO) efforts to build support for APIs like GCE, CIMI or
OCCI.

Cheers,
Mark.




More information about the OpenStack-TC mailing list