Time-based updates are right for feel and wrong for rules that must agree across machines. The fixed-step machine runs the rules in equal slices and carries the leftover time, so the same inputs produce the same states on any machine.
About 1 min of reading
Convert a loop to fixed steps
Write the accumulator
Before you run it, say out loud what it will print. Then open the answer.
Run whole slices
Before you run it, say out loud what it will print. Then open the answer.
Prove the same states
Before you run it, say out loud what it will print. Then open the answer.
Test the cap
Before you run it, say out loud what it will print. Then open the answer.
Before you write anything, put these lines back in the right order. Click a line to move it up one step. This support step is easier than writing code from scratch, and it is here to make the last example stick.
A slice shorter than the fastest movement still catches collisions. A slice longer than a wall thickness tunnels, so the slice is a property of the level geometry.
Presentation only: particles, camera smoothing, animation. Those read the leftover time as a fraction and never affect the state.
The rules always receive the same dt. The number of calls varies with the machine, the values per call do not.
Accumulate the elapsed time and run steps while there is a whole slice left. The leftover stays for the next frame, so time is never lost.
A long stall would run hundreds of steps to catch up. Cap the steps per frame and drop the rest, so a stall is a hitch and not a freeze.
Each one is defined here on this site. Open the entry for the full definition and the other words it points at.