Source Location for EJB3 Components

[alrubinger]

Persuant to discussions Carlo, Jaikiran and I have been having lately, I've extracted the component charged with bringing @Asynchronous support to EJB into its own space:

http://anonsvn.jboss.org/repos/jbossas/projects/ejb3/components/

Under this alignment "org.jboss.ejb3.async" is a component with "spi", "build", and "impl" *modules*.  Each module shares a release cycle with the component as a whole.

Criteria:

1) Each component has an SPI upon which other components can depend

2) In compile scope, no component may depend on anything aside from the designated SPI.  In runtime/test scope the doors are open to use impl classes (I *really* wish Maven provided a scope for "test runtime only, but not test compilation".

How do we feel about using this setup as a template for further breaking apart EJB3 pieces under trunk?

S,

ALR

[alrubinger]

For better or for worse, I tracked these commits under https://jira.jboss.org/jira/browse/EJBTHREE-1721

S,

ALR

[jaikiran]

ALRubinger wrote:

2) In compile scope, no component may depend on anything aside from the designated SPI.

I can't seem to understand this one Can you elaborate? For example, the async component is composed of multiple modules. So are we saying that the individual modules should not depend on anything other than the SPI?

[alrubinger]

jaikiran wrote:

I can't seem to understand this one Can you elaborate? For example, the async component is composed of multiple modules. So are we saying that the individual modules should not depend on anything other than the SPI?

Sure.  In this case ejb3-core is probably the only consumer of ejb3-async.  So it'd depend upon ejb3-async-spi in compile scope, while the test runtime would need ejb3-async-impl as well.  This ensures that components don't rely upon internals; when we need to do refactoring and bug fixes, we can limit the impact.

S,

ALR

[jaikiran]

ALRubinger wrote:

Sure.  In this case ejb3-core is probably the only consumer of ejb3-async.  So it'd depend upon ejb3-async-spi in compile scope, while the test runtime would need ejb3-async-impl as well.

That makes sense. So it's actually the consumers of the component, which shouldn't depend on anything other than the spi in compile scope.

[alrubinger]

Minor change after a discussion w/ Carlo:

I've moved the structure to leave space representing branches for each component.

http://anonsvn.jboss.org/repos/jbossas/projects/ejb3/components/async/trunk

S,

ALR