On Fri, May 30, 2008 at 5:53 AM, Brian E. Fox <[EMAIL PROTECTED]> wrote:
> I personally think we have conflicting rules in the way we handle > incubator releases. > > > > On the one hand, we require incubator releases to be in a separate > repository... for whatever reason (they aren't part of Apache, they > aren't stable enough, etc). On the other hand, we allow regular releases > of apache artifacts into the central repository with dependencies on > incubator artifacts. > > > > On many occasions, I've seen this cause people lots of confusion because > they update to a new version of an existing artifact and suddenly their > build fails to find a dependency. (because the new version is now using > an incubating artifact) > > > > IMO, things going into the central repository must have their entire > transitive hull available in the central repository. Therefore, we must > draw one of two conclusions: > > > > 1. Incubator releases go into Central > > 2. Regular releases cannot use Incubator artifacts > > > > Since the whole point of the incubator releases is to get some people to > use them and prove them out, I say 2 is not really an option. If the PMC > of a given project tests out an incubator artifact and deems it good > enough for a release, then that should be enough.... But let's make it > easier for the users by having those dependencies available. #2 would also present a big issue for any existing incubator project that depends on another project. What happens if my project graduates by my dependency does not? Should incubator projects not depend on other incubator projects as well? > > > > --Brian > > > >