DI is a module system built out of classes; dynamic scoping is another way to build a module system. If you had a good composable (parameterized) module system

DI is a module system built out of classes; dynamic scoping is another way to build a module system. If you had a good composable (parameterized) module system, you'd have much less need of DI. A composable module system would scope the lookup of type names and static methods to the actual module arguments arguments the accessing module is constructed with. The problem with `new X()` vs `@Inject X x` is that the construction of X in the former has no indirection; type names are global constants. A module system provides an indirection. Dynamic scoping could also provide an indirection, because dynamic scoping lets you redefine / redirect those otherwise constant things. (DI in practice does a bunch more, like proxies to let you put data with different lifetimes (request, session) in a mixed object graph; and the fact that proxies now exist means aspect-oriented programming sticks its head in and encourages its use for things like auth and transactions. Once you go over the edge of the DI barrier to acceptance, "best practices" shift dramatically - you end up quite far from where you started.)