dev@jsr311.java.net

Re: API feedback

From: Dhanji R. Prasanna <dhanji_at_gmail.com>
Date: Fri, 13 Jul 2007 05:17:19 +1000

On 7/13/07, Marc Hadley <Marc.Hadley_at_sun.com> wrote:
>
> On Jul 11, 2007, at 6:37 PM, Dhanji R. Prasanna wrote:
>
> >
> > This is something I brought up very early on but would tend to
> > discourage as we cannot provide a coherent implementation of
> > @Resource without confusing the hell out of developers working in
> > SE environments or (future) EE profiles that are not of the
> > "kitchen sink" ilk.
> > It may be worthwhile to support it as an "also possible" in certain
> > EE profiles (enabled with jndi registries).
> >
> I note that @Resource is included in SE 6, is your concern that
> @Resource will only work with certain types in certain contexts/
> profiles ? If so I think we can counter that problem by being very
> clear about what is guaranteed to work where.


Yea this is essentially it. Right now any servlet container without jndi
cannot do justice to @Resource.
I imagine this will iron itself out as the various other JSRs struggle with
IoC too.

> - Resource injection needs to happen in one place. JAX-RS, Servlet,
> > JSF, JAX-WS, WebBeans, Persistence, EJB, ... all perform resource
> > injection and its possible to mix annotations from multiple
> > technologies in a single class. It would be useful to have some
> > common infrastructure across EE for performing the injection, perhaps
> > coupled with infrastructure for instance creation and management.
> >
> > Aha! This is what Ive been trying to say, somewhat less than
> > eloquently (apparently!) in other threads of late. I think this
> > needs careful consideration and leadership from either JSF2 and Web
> > beans as their IoC concerns will be the most robust and wide-
> > ranging. We would also need careful coordination with the Umbrella
> > EE JSR to make everything happen.
> >
> FWIW, I think JAX-RS has the same need for robust and wide-ranging
> IoC. The key thing (IMO) is to have some consistency across the
> platform specs. E.g. it seems reasonable for a JAX-RS resource class
> to be able to benefit from injection of Java persistence artifacts,
> EJB references etc.


Well, I meant that the driver for IoC is much stronger and more wide-ranging
in Web Beans and JSF2 (Conversation scopes, 2-way injection, staged
lifecycle etc.). I agree, however, that we ought to play on the same field.

>
> Sounds good, perhaps around the time we publish an early review
> draft. That would be a good opportunity to get feedback.


Good stuff.

Dhanji.