The future of X, if it has one, is to rip the backend off a Wayland compositor, deprecate most legacy features (server side fonts, most drawing primitives) and

The future of X, if it has one, is to rip the backend off a Wayland compositor, deprecate most legacy features (server side fonts, most drawing primitives) and most visuals (e.g. Xrender has a required set of visuals; support those and no others) then test app by app you care about and add client side shims of there are any apps anyone cares about that can't be patched. Most X apps these days use only a tiny subset of request types anyway (source: lots of time spent with xtruss) and most of what's left is easy to shim. The X protocol is messy, but very little of it is used by software people both a) care about, and that b) isn't available in source form and can easily be upgraded. Sometimes I wonder about taking wlroots and doing a FrankenCompositor with first class X protocol support, including window management and all the bits the Wayland proponents keep not wanting to add...