-----Urspr�ngliche Nachricht-----
>Von: Rickard �berg [mailto:[EMAIL PROTECTED]]
>Gesendet: Freitag, 8. Dezember 2000 09:36
>An: jBoss Developer
>Betreff: Re: AW: AW: [jBoss-Dev] Proposed refactoring of
>ContainerFactory -Ch ainingDeployment services in general
>Are you saying that you are having multiple EAR applications, and where
>an EAR can optionally specify it's parent EAR?
well, yes. I should have used more J2EE vocabulary in order to make
it clear. Whenever referring to "ejb-jar", I really meant "EAR application"
- maybe confused
that because we are currently not using JSP,
but a thick and comfortable VB-GUI.
>The biggest reason seems
>to be to allow classloader sharing, which is only relevant for call
>optimization (no serialization needed). Is that a correct understanding
>of what you're saying?
>But multiple EAR's would do the trick, right?
>But what if sales and stock both references masterdata *jar*'s through
>classpath manifest? I.e. they are seemingly redundant but in reality use
>the same jar. Would that be ok?
My argument wrt minimising redundancies would be addressed by this
technique.
The only really relevant argument remains ... performance and optimisation
argument, yes.
>If I understand you correctly, all you're saying is that you want an EAR
>to be able to specify a "parent", and then that parents classloader
>should be used as parent to the EAR's own classloader. Right?
You made a long talk very very short (I bet I�ll get a nick like
CG"verbose"J on this list). You were an analyst in your previous life,
were�nt you ?
>Nevertheless, yes this is interesting :-)
puuh, interesting enough to extend the J2EEDeployer (undermine, undermine
...) ?
I think this would not violate the specs, would it?
CGJ
