Sectus
[13] Hero
I'll just reduce my argument to my main point: I think it would be more worthwhile to petition Namco into making a better netcode at all, GGPO shouldn't be the only option. Shooters have found a ton of different solutions to latency problems, fighting games shouldn't be limited to just 2 solutions (one of them being nearly useless, the one which most games use right now)
The type of netcode which most fighters use right now has the benefit of being very easy to implement and it should always stay in sync (same netcode which SC4 used). It works in the way that each client (client being each running version of the game) initially is paused and only processes a frame once it has received player input from all players. Once it starts running the game it should run smoothly, but it always processes old player input. The input lag becomes roughly the same as the ping. This should mean 50-100ms if you play against someone who are in the same country(this is playable, but it's still enough input lag to make it too different from local play), or more than 100-200ms if you play against someone who's in another country (this is where it starts to become close to unplayable).
GGPO works in a different way. Instead of waiting for player input, it'll do a form of client side prediction. Once you do any input, it'll process them and show the effect on screen, and the opponent's inputs will be processed as they received. Whenever there's some kind of desync (for instance, you attack the opponent and he gets hits, but on his screen he was able to block in time), then it'll roll back to a previous frame which was in sync, it'll quickly re-process the previous frames with up to date player input and then resume at the current frame with the new proper state of the game. GGPO also adds a small constant amount of input lag to reduce the chance of desync (I think this is 20ms or less).
GGPO has the downside that you'll need to reserve memory for many previous frames for rollbacks (although this should only be a concern for emulated games) and that you need to be able to re-process frames in the timespan of a single frame (otherwise the game might freeze for a split second whenever it rollbacks).
I may have a few details wrong about GGPO, but that's how it works overall. The upsides are pretty obvious: you'll have far far less input lag. It also works as an overall netcode solution, you shouldn't have to add lots of code for special cases. It's also one of the few netcodes which works with emulators (you're extremely limited when it comes to adding netcode to an emulated game).
As a quick history lesson: the former netcode I mention was very common for all games in the 90's, but then most games were played on LAN at which case you were dealing with so low pings it didn't matter. That kind of netcode is pretty much obsolete now. Outside of fighting games and emulated games, you practically never see it.
I can give a quick explanation for the 2 netcodes (I can't go into detail about VF5, as I know nothing about it).
The type of netcode which most fighters use right now has the benefit of being very easy to implement and it should always stay in sync (same netcode which SC4 used). It works in the way that each client (client being each running version of the game) initially is paused and only processes a frame once it has received player input from all players. Once it starts running the game it should run smoothly, but it always processes old player input. The input lag becomes roughly the same as the ping. This should mean 50-100ms if you play against someone who are in the same country(this is playable, but it's still enough input lag to make it too different from local play), or more than 100-200ms if you play against someone who's in another country (this is where it starts to become close to unplayable).
GGPO works in a different way. Instead of waiting for player input, it'll do a form of client side prediction. Once you do any input, it'll process them and show the effect on screen, and the opponent's inputs will be processed as they received. Whenever there's some kind of desync (for instance, you attack the opponent and he gets hits, but on his screen he was able to block in time), then it'll roll back to a previous frame which was in sync, it'll quickly re-process the previous frames with up to date player input and then resume at the current frame with the new proper state of the game. GGPO also adds a small constant amount of input lag to reduce the chance of desync (I think this is 20ms or less).
GGPO has the downside that you'll need to reserve memory for many previous frames for rollbacks (although this should only be a concern for emulated games) and that you need to be able to re-process frames in the timespan of a single frame (otherwise the game might freeze for a split second whenever it rollbacks).
I may have a few details wrong about GGPO, but that's how it works overall. The upsides are pretty obvious: you'll have far far less input lag. It also works as an overall netcode solution, you shouldn't have to add lots of code for special cases. It's also one of the few netcodes which works with emulators (you're extremely limited when it comes to adding netcode to an emulated game).
As a quick history lesson: the former netcode I mention was very common for all games in the 90's, but then most games were played on LAN at which case you were dealing with so low pings it didn't matter. That kind of netcode is pretty much obsolete now. Outside of fighting games and emulated games, you practically never see it.
