05:42 <leon-p> ifreund: I agree with the proposal, I can't think of any downsides and it seems to solve the issue cleanly

05:47 <leon-p> I'll likely have to add some safe-guards so that lisp code donig it in the wrong order will raise an exception rather than commiting a protocol error causing the program to exit, but I want to redo error states anyway

07:33 <ifreund> leon-p: cool, since writing it up the only downside I've come up with is that it adds one extra roundtrip of latency when the server is idle and the window manager initiates a state change

07:34 <ifreund> I think I can live with that though I will give it a bit more thought to make sure I can't fix that without sacrificing the simplicity

15:00 <leon-p> It should be possible to signal the server to skip one roundtrip, but honestly I don't think it's worth the extra complexity

15:01 <leon-p> the WM doesn't perform enough actions often enough for the latency to ever be noticeable or relevant

15:01 <leon-p> (at least any sane WM)

15:03 <leon-p> my guess is that averaged over a work session, the average WM roundtrips per minute will be lower than 100 (majority of that likely keybinds and focus changes)

15:09 <leon-p> any non-trivial Wayland client implementing a user interface via xdg-shell will have a higher amount of roundtrips (especially anything that has any kind of text input) so it should be perfectly fine

15:10 <leon-p> and even should we decide to add high-freq. events to the WM protocol, like mouse position, we can always make them opt-in (i.e. requiring a request to activate or sending them via some object that needs to be bound explicitly)