[rhetorical] Should AttributeResolver implement the Resolver interface ?

[Rod Widdowson]

> My suggestion is : if we have vague agreement on
> getServiceableComponent() then that should be decision enough to wire
> up the AttributeResolver(Service) into java-idp-testbed. Bits and
> pieces from the unit tests should be enough, I think.
> 
> Then talk to us more about the refcounted model.

Kinda-sorta.  

If we are going to have a getServiceableComponent() paradigm then we must
either have a reference count or not support reloading (or not support
destroyable components)

So to my mind a vague agreement on getServiceableComponent( ) is a vague
agreement on reference counting.  

But since this really isn't a huge amount of work, we can certainly delay
discussion about recounting until we are half way through the
implementation.  If we do decide to rip it out it's not a huge deal.

> Maybe (0) leave the AttributeResolver+Service the way it is now, (1)
> wire that up into java-idp-testbed, then (2) work on
> getServiceableComponent, then (3) refcounts.

Sure, makes sense (modulo the above).

R