admin@glassfish.java.net

Re: follow up on last Friday's SAF/AMF discussions

From: Hong Zhang <Hong.Zhang_at_Sun.COM>
Date: Mon, 24 Sep 2007 13:09:35 -0400

Hi, Sreeni
>> I've followed up with Binod and confirmed that AMF does not need to
>> manage standalone appclient module. :-) Nandini, so please update the
>> one-pager by removing the requirement on appclient under jsr77
>> implementation section.
>>
>> It seems the AMF does need to manage the resource adapter module, so
>> that item will remain on the list.
> I had sent an email to Binod explaining the jsr77 role in handling
> resource modules.
I don't think the spec mandates how the implementation of the start/stop
should be. The current implementation that we have for resource adapter
is not wrong. And Binod's proposal to modify the implementation of
resource adapter is for the purpose of enabling the AMF use case.

>>
>> I also did some investigation on why the JSR77 start did not load the
>> previously deployed (with disabled state) ear application during
>> server start up. It seems in the enable path we have a few places
>> checking for "enable" attribute before actually loading the
>> application, and the "enabled=false" prevent the application from
>> being loaded. So we might have to loosen the checks in those places
>> for this to happen.
> I think we still need to honor the value for enable attribute, for ex.
> if it is false then the app. should not be loaded. We need to make
> sure that the jsr77 start/stop works fine.
Yes, we need to be careful about making changes in this area. We should
not cause any behavior regression, and we also need to make the JSR77
start work when the enabled is false.

Thanks,

- Hong