Hi,
On 12/20/2012 01:21 PM, Sivakumar Thyagarajan wrote:
>> 2) Yes, application clients should be excluded
>
> I went back and checked the EE spec and our spec, and it appears that we
> don't /require/ support for Connectors in application clients, but
> vendors may choose to do so. So, we cannot exclude it, but we don't need
> to require support for the resource definition annotations in
> application clients.
>
Yeah, that is what I meant. So ok.
>> 5) Deployments to component namespaces (java:global, ...) should be
>> excluded in this revision.
>
> I assume you want this restriction only for resource definition
> annotations placed in resource adapter modules. Do you have a particular
> reason for this restriction? What would be the use if we don't allow a
> registration to any component namespace.
>
Yes, even though :global and :app would be "easy" to add in this
revision I believe that we should leave the entire JNDI subject for the
next revision where we can discuss the matter in detail in the time
frame it requires.
I also believe that it would benefit deployments of resource adapters to
have default JNDI binding for the cases where we could support this.
That area also would need to align with the component namespace discussion.
So it wouldn't be good to make spec changes now that turns out not to be
optimal down the road.
A vendor can choose to support component namespace rules outside of the
spec.
>> We need to exclude component namespaces in this revision such that we
>> aren't restricted once we start to work on standardizing JNDI locations
>> and component namespaces. Once these policies are in place we can update
>> the policies for the annotations based on the outcome.
>
> I agree that the lack of a component namespace for resource adapters is
> a problem, and hence allowing a resource definition annotation in
> resource adapters to place resources in java:comp or java:module is a
> challenge.
> One alternative is to only allow resource adapters to create
> java:app or java:global resources. This works and is okay until we
> define a component namespace for RAs, but is it the best approach? I
> would love to hear from you and others in the EG on this.
My comment is as above.
Best regards,
Jesper