To best handle multi-tenant and/or flexibility on context dependencies, its necessary implement on graph context the following needs:

1. a graph context can have several names (an array) - each name should still be unique, so you can attach several unique names to on context. This is important to help identify a context by soe internal name (like a sha1 of a jar file) and a more conventional name like: osl-common-1.0.jar
2. contexts should be able to group other contexts, its important to simplify some king of configurations. A good example of this need is about jre-6 that will just point to other contexts like rt.jar, ui.jar and etc.. the same happens for jee and etc..