Have you guys ever thought about business benefits from the module descriptor?
If the module descriptors exist, they have to be tested.
Unless we have a business feature with module descriptors in Jakarta EE, it is useless to introduce them in the JAR.
I have a business feature in my mind:
Imagining that you have EE containers in a Cloud. Your application has module descriptors and the application can select the EE stack and versions. Then your microservice would not include the infrastructure JARs because these JARs/modules can be selected via the module descriptor of the application. The trick is that the Fat JAR would not include the infrastructure JARs because they are on the server and the best is that the application can select the version of the API and container. I am calling it software as service on the Cloud, and the s/w service is a list of containers and their vendors.
After we have provided the ability to select the version of the Spec in the application, we would kick off a new strategy of the deployment (small JAR/WAR without embedding container jars in the app) and new services in the Cloud.
It is something between Fat JARs and application servers because your app is really server-less and it is not like an application server but the Cloud vendor would have a container of containers.
T
