I think the goal in making sure that the directory structure corresponded to pom structure (in terms of parents) was more about making it easier for humans to understand and parse the the variety of poms in a project easily.

Maybe the *right* approach is to put a comment in the actual root (aggregator) pom saying that "this is not the logical parent, in pom space, the logical parent can be found at: <path-to-project-parent-pom>". Alternately, if there are a few different parent poms, we should explain that there. Ideally, we'd also communicate to all projects (both as tutorials for projects and as best-practices to follow going forward) what a common-case setup would look like.

If projects mostly followed these patterns it would help a lot.

--Colin

On Wed, Oct 15, 2014 at 1:00 PM, Robert Varga <nite@...> wrote:

I think we need to look at what our current projects with nested artifacts do and why. In BGPCEP we tried defining things in parent directory poms, and it does not work out that great. It actually creates more places where things are (re)defined, making debugging more difficult -- and saves exactly a single line of XML when compared to 'all artifacts point to project-wide parent', two if your subdirectory comes with a different version than your project-wide thing is.

Thanks,
Robert