[webbeans-dev] Web beans implementation plan

[Pete Muir]

On 20 Oct 2008, at 17:28, Christian Bauer wrote:

>
> On Oct 16, 2008, at 22:55 , Gavin King wrote:
>
>> We now really need to get going on the RI. I've attached the latest  
>> draft.
>
> Comments up to chapter 2 - I might be completely wrong with my  
> questions because of something that I might find in a later chapter,  
> so ignore if that is the case:
>
> Chapter 1
>
> @Model is not used on model components in the example, they are  
> actually controllers and it probably should be called @Action. Or  
> you need to write different examples that actually put @Model on the  
> model.
>
> (Minor) Remove the System.out.println("Well, read the exception  
> name...") and re-throwing exception pattern in the examples. It's  
> embarrassing and if even one user copies that pattern, you've made  
> many peoples lives more difficult. Either log, or re-throw, never  
> both.
>
> I like the decorators, great stuff.
>
> Chapter 2
>
> Seems all good, I like what I'm seeing with XML namespaces in the  
> descriptor. A first for Sun?
>
> 2.5.3: I can easily imagine a component I want deployed in two  
> scenarios, so we should allow multiple deployment type declarations  
> on a web bean or producer method and only validate the highest  
> precedence.
>
> 2.5.7: What's the value in enforcing that the <Standard/> deployment  
> type always be declared? Just enable it.

No, it's much neater doing it this way.

>
>
> 2.6: I understand that nobody cares about EJB names but why can't it  
> be a web beans name? At least it would be useful for something.
>
> 2.7: Inheritance of stereotypes is probably not needed. I can't see  
> an application having more than a few stereotypes and 2.7.2 already  
> defines composition of stereotypes.
>
> 2.7.5: Going back to my first comment about chapter 1, I recommend  
> removing the built-in @Model stereotype. I have no idea what I would  
> ever use it for. A typical conversational MVC application certainly  
> has not many request-scoped model instances...

It's gone, this must be leftover from an old example.

>
>
> 2.8: The whole section and functionality of explicit specialization  
> seems unnecessary. Why can't I just put @Inherited on my binding  
> type declaration? If, in the example, I would put @Inherited on the  
> declaration of @Asynchronous, the same specialization rules for  
> binding and naming could be derived.