[jsr342-experts] Re: CDI positioning

From: Bill Shannon <bill.shannon_at_oracle.com>
Date: Mon, 05 Nov 2012 16:09:58 -0800

Based on the feedback I've received on this issue, it's pretty clear that
there's very strong support for CDI, and yet sympathy for those who want
their specs to work outside of Java EE without requiring CDI.

Pete even proposed this position, which some of you supported:

> D. Technologies that can be standalone specifications (JMS, JAX-RS) should be
> fully functional without CDI, but they should not duplicate any features of
> CDI. When CDI is present, these technologies should leverage and integrate
> with CDI where appropriate.

This seems similar to my "C" below, but with only a soft dependency on CDI
when used standalone.

This seems to mean that we should accommodate those specs that want to
work without CDI, even though we all love CDI! I have a specific question
in this area in my next message...

Bill Shannon wrote on 08/30/12 13:58:
> From many of our recent discussions, it seems clear that CDI is
> becoming more central to the Java EE programming model. For example:
> - The expanded use of @Stereotype in my previous message.
> - The use of CDI interceptors to provide container managed
> transaction support beyond EJB.
> - The potential future use of CDI interceptors to provide container
> managed security support beyond EJB.
> - The use of CDI interceptors to support Bean Validation method
> level validation.
> - The discussion of "implicit producers" to allow use of @Inject
> instead of @Resource to inject Java EE resources.
> - The discussion around alignment of CDI managed beans and the
> separate @ManagedBean spec.
> - The introduction of a transaction scope and its use in the JMS
> spec to simplify the programming model.
> - The change being considered by the CDI expert group to enable
> CDI by default, making it more attractive to use it for all
> the items above.
> At the same time we're finding that some specs, e.g., JAX-RS, are
> hesitant to introduce a hard, or even soft, dependency on CDI,
> instead insisting that all their new features must work in an
> environment where there is no CDI.
> In many ways this parallels what we saw with annotations. In
> the beginning we found many people who didn't want to use annotations
> and wanted us to make sure everything worked without use of
> annotations. Now we're seeing many things that *only* work with
> annotations, and annotations are well accepted by Java EE developers.
> I suppose there's a natural lifecycle to acceptance of new
> technologies, and I wonder where we are in that lifecycle with CDI?
> Has CDI become a mature and accepted technology that we should use
> widely?
> I'd like to get a sense from this group as to what direction we
> should provide to all the Java EE specs in regards to their use
> of CDI. Here's a few obvious options:
> A. Technologies that see a significant standalone (non-Java EE) use
> should be fully functional without CDI. If necessary, any
> required features that are similar to CDI features should be
> defined and implemented in a way that doesn't depend on CDI.
> B. Technologies should provide all major features in a way that
> works without CDI. Some features may also be provided in a
> different way that works well with CDI. Some less essential
> features may only work with CDI. The implementation should
> only have a soft dependency on CDI at most.
> C. Technologies should provide features that work well with CDI
> without duplicating any functionality in CDI. Use CDI wherever
> it fits. The implementation may have a hard dependency on CDI
> and may require CDI even when used in a standalone environment.
> I'm sure you can think of other options as well.
> What advice do you think we should give to other Java EE specs?