[jsr342-experts] Re: CDI in Application Client Container

From: Bill Shannon <bill.shannon_at_oracle.com>
Date: Fri, 16 Nov 2012 15:20:43 -0800

Jason Greene wrote on 11/16/12 13:09:
> On Nov 16, 2012, at 3:06 PM, Bill Shannon <bill.shannon_at_oracle.com> wrote:
>> Pete Muir wrote on 11/15/12 03:54:
>>> Stuart Douglas just raised these concerns as well:
>>> * Which contexts should be active: I would say just the application context,
>>> which is probably not hugely useful (this is the same point Nigel raised).
>> Maybe not all that useful, but well defined and easy to define.
>> We could define that the entire application runs in a single request scope,
>> which would help JMS 2.0.
>> We could also define that each event processed by the AWT event dispatch
>> thread is a separate request scope, but that seems excessive.
>>> * The other problem that just occurred to me [Stuart] is that local EJB's are
>>> not present in the application client container, so if you have some code
>>> that is mean't to run server side, the presence of that code will cause an
>>> unsatisfied dep error. In fact the more I think about it the more problematic
>>> this seems.
>> I don't understand this one.
> There's an expectation that you can add EJB capabilities to a CDI bean by adding annotations, that would not work unless we added local ejb
> support to ACC. Although I don't see why we couldn't add support for that. It could be useful. I mean the way I have always viewed ACC, is just a trimmed down server that can act as a client.

We're not planning to add EJB support to ACC, but note that we're
refactoring many of the EJB capabilities so that they can be used
more widely, such as interceptors in EE 6, transactions in EE 7
(yes, there's no JTA support in ACC so this doesn't help), and
possibly security in EE 8.

What EJB capabilities would people like to see in the ACC?