@nex said in Why would you have left FAF?:
Supcom has an inbuilt input latency, which is independent from both the sim latency(which comes on top, if at all) and network latency(which the input latency is meant to compensate for).
so no matter how good your ping is you will have those 500ms of input lag. this could in theory be a variable lag, but from what I can remember it's fixed (and apparently to 500ms), to make every game feel consistent.
so every tick (or possibly every frame) the game collects your inputs for tick n+5 and sends them to all other players.
At tick n+5 the game checks if it received inputs from all players and if not pauses.
So these 500 ms are currently fix. assuming the game might take about 100ms to send your inputs(if it gathers them all over the course of a tick and then sends them once) then you get up to 100 ms of extra lag (if you press right before it's send you get 0ms, if you press right after a send you get 100 ms extra lag), so on average that is 50 ms lag, which you could halve by doubling the sim rate, so you'd gain 25ms on average but the 500ms (which is chosen so that players with a ping of up to 250ms to each other can play together) will always come on top of that, so the net gain is between 5-10%so tl;dr there is a 500ms inbuilt lag to compensate for any network lag, which overshadows any amount of lag introduced by the sim tick rate.
Yes, I talked with Jip yesterday and he confirmed that inputs are scheduled for 5 ticks in the future. It's possible to (almost) completely eliminate this by scheduling the inputs at a proxy server like sc2 (or seemingly BAR). Basically they asymmetrically schedule inputs based on latency, such that people with high ping to the server have higher input lag than those with low ping to the server. So in those cases input lag is totally dominated by sim tick length.
I get what you are saying now though, I thought you were saying that there's some nebulous other source of input lag or that network latency dominates, which is not true. It's just a somewhat arbitrary value of 5 ticks chosen to be acceptable by gpg in order to avoid writing dynamic scheduling logic x) Even in pure p2p it should be easily possible to communicate input scheduling changes based on latency (though then the scheduling has to be synchronous).