"alesj" wrote : | ..., since demand/supply concept handles things beyond contexts handled by MC - a generic match mechanism. And as such some items we demand might not have dependent state. | This part is wrong, or at least it doesn't explain the real problems behind our demand/supply. First, items *do* have dependent state, since every supply is backed-up by a kernel controller context. It's the way we have it implemented. Currently we add/register all supplies from BeanMetaData at Installed state. So no way of knowing if some supplier can supply the supply before (a lot of supplies here :-). We can extend the supply notion in a similar way we extended installs, to be applicable in any state. Then having | Map<ControllerState, Map<Object, List<KernelControllerContext>>> | in AbstractKernelController. "alesj" wrote : | Should we create new context-demand notion? | Definitely not. :-) View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4152544#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...