Could you elaborate "fold the screenreader into the compositor"? The thread that forwarded Magdalen got somewhat messy after several answers, but as far as I understood, we never proposed to move the screenreader to the compositor. We were just debating about some functionality/info that we got from X, and that now we need to get from somewhere else (Wayland or the compositor).As I've said to Joanie and API in Montreal, we probably need a privileged Wayland interface for ATs to obtain global information like the pointer position and the surface under the pointer. This will be needed to support some of orcas functionality. And I don't think it makes sense to fold the screenreader into the compositor.