The Problem is ...
Most likely the pathfinding routine. If you, during your turn, do something that requires the pathfinder to run, then there is a lag created. But, because you move at human speeds, the pathfinding lags aren't really noticeable overall. When the AI takes there turn, esp. if there is a war, the lag becomes very noticable (because you see it all at once). Now add MP to this lag; the communications IP overhead, the Type of cummunication (33K to ADSL/Cable), the Server state (light to heavy load) AND the Pathfinder being run on all connected machines and what would "likely" happen would be even slower than it is in SP. Even different screen sizes would have an effect.
The "best case" would be to not have ANY AI's play in the Game, but still you will have to contend with any/all associated Pathfinder delays while in the process of doing something else. And what about the Barbs?
What's the "cure" for this?
A Client/Server version that would have the Meat of the Pathfinder (and the game) on the server machine and just have the Client mostly used as a game terminal? That would really create a problem if the server has several games going at once.
Or how about a distributed computer approach to the MP/Pathfinder problem? But then how do you reconcile the different machines connected at different speeds issue? Split the work up in proportion to these things, but what happens is someone drops out? Do you start from scatch if that happens?
I think that the best playable solution is some form of middle ground between the two. If you make some form of change within your view area (non-FOW), then your machine should do the pathfinder calcs for that section only, then relay the results to the server. If the applicable path requires going into the FOW, then the Server should determine whos machine would run that part of the pathfinder. The server would also be responsible for all Mutual players FOW pathfinder. And during the AI turns, if any, break up the pathfiner between all of the players machines.
Just a thought on why MP hasn't been done yet.
Maybe, the pathfinder needs to be more intuitive, i.e. don't worry about computing a lot of paths if only a few are affected by player actions. I think that Fraxis should Open Source it's pathfinder Algorythm, so that true MP pathfinding could be hashed out. This might also have an effect of speeding up the SP game, if a more code efficient (robust) approach is thought of.
E_T
Most likely the pathfinding routine. If you, during your turn, do something that requires the pathfinder to run, then there is a lag created. But, because you move at human speeds, the pathfinding lags aren't really noticeable overall. When the AI takes there turn, esp. if there is a war, the lag becomes very noticable (because you see it all at once). Now add MP to this lag; the communications IP overhead, the Type of cummunication (33K to ADSL/Cable), the Server state (light to heavy load) AND the Pathfinder being run on all connected machines and what would "likely" happen would be even slower than it is in SP. Even different screen sizes would have an effect.
The "best case" would be to not have ANY AI's play in the Game, but still you will have to contend with any/all associated Pathfinder delays while in the process of doing something else. And what about the Barbs?
What's the "cure" for this?
A Client/Server version that would have the Meat of the Pathfinder (and the game) on the server machine and just have the Client mostly used as a game terminal? That would really create a problem if the server has several games going at once.
Or how about a distributed computer approach to the MP/Pathfinder problem? But then how do you reconcile the different machines connected at different speeds issue? Split the work up in proportion to these things, but what happens is someone drops out? Do you start from scatch if that happens?
I think that the best playable solution is some form of middle ground between the two. If you make some form of change within your view area (non-FOW), then your machine should do the pathfinder calcs for that section only, then relay the results to the server. If the applicable path requires going into the FOW, then the Server should determine whos machine would run that part of the pathfinder. The server would also be responsible for all Mutual players FOW pathfinder. And during the AI turns, if any, break up the pathfiner between all of the players machines.
Just a thought on why MP hasn't been done yet.
Maybe, the pathfinder needs to be more intuitive, i.e. don't worry about computing a lot of paths if only a few are affected by player actions. I think that Fraxis should Open Source it's pathfinder Algorythm, so that true MP pathfinding could be hashed out. This might also have an effect of speeding up the SP game, if a more code efficient (robust) approach is thought of.
E_T
Comment