From the wayland readme:

> The wayland protocol is essentially only about input handling and buffer management

So sending input to clients and managing the buffers needed is what wayland implements. Everything else is either done in the clients (e.g. rendering) or the compositor (pretty much everything else).

Note that this is a subset of what X11 did (though rendering had pretty much already moved to the client for modern applications). X11, of course, handled input, and it also provided an IPC mechanism (which is how clients e.g. communicated things to the WM). Wayland, at its core, is just those two things.