What about a mix: Every 100ms you fit the best Beizer curve to the last batch of mouse positions and send that. It seems like that would give a more precise r

What about a mix: Every 100ms you fit the best Beizer curve to the last batch of mouse positions and send that. It seems like that would give a more precise reconstruction than fitting a Beizer curve only to one sample every 100ms on the server.