There are some statements on the Project Atomic website about Nulecule. There is also some information in the Nulecule Github repository. I've spent big portion of DockerCon explaining Nulecule to various people and here is what my explanation boiled down to:
As a deployer of multi-container application, I want to have simple way how to parameterize orchestration manifests
- f.e. if you look at kubernetes/examples, you can see "change this, and that in the JSON configs to make this app working in your environment" in every other example.
As a developer of a multi-container application I want to use existing components (like databases) in my application so that I don't have to burn my time on something that already exists.
- f.e. as someone creating a voting app (borrowing from DockerCon:) ) I want to use Redis and Postgresql components and only add frontend and worker to the equation.
3. Multiple Orchestrations - providers
As a developer of a multi-container application I want enable deployment to multiple orchestration providers for my application so that users can easily migrate.
- f.e. I start on Deis, but decide GiantSwarm is better, than I go nuts and buy OpenShift Enterprise and easily migrate there, though things go south and I go to a plain CentOS + Kubernetes to save some money...
As an enterprise consumer of a multi-container application I want to avoid any out-of-band transport layer for the application and it's metadata
- f.e. I use Docker images and have a private registry set up. Instead of having to setup another authenticated webserver and figure out how to verify what I pull as tarballs or plain text, I will package every piece of puzzle into a Docker image and distribute it through registry.
The next and very valid question is: "How well did we tackled down these challenges?" That's up to you to figure out and tell us and ideally help us fix bits that you struggle with.