I''ve never written a multi-player game, but I would like to implement that aspect into my current project in the future. My current project is in the design/baby stage, but it is a good time to start thinking about networking issues. Based on a Client/Server architecture, I am thinking about having something along the lines of the following: Single player mode mimics client/server architecture; the server just resides on the user''s machine. Create an ObjectManager that stores a list of all rendered objects in the game. The ObjectManager resides on the server. Each rendered object is responsible for rendering itself - object.render(); The client machine updates itself(basically input, Camera Movement). Calls object.update() for each object it directly controls (playar avatar), which sends update info to the server. Server calls object.update() for each object it controls (any non-static geometry, projectiles, particles, etc). Client calls render(), which gets a list of renderable objects from the Server, which then render themselves. Basic premise, good or bad? Some questions: Should the client keep a copy of the Object Manager, that is updated based on the server''s ObjectManager? This would prevent updating things that weren''t moving, and would only have to send info concering updating the changed items rahter than sending render object b that hasn''t moved). Scalability? This seems like it would work well for your basic FPS, 15-20 players. But for a MMOG. with 1000+? Seems like it would tank really hard. Creation and Control. Say your basic FPS. Players A, B, C. Player A throws a grenade. As soon as it is thrown, the server gets control of it. Should client for player A get updates concerning this grenade, ot should A already know that it created the grenade, threw it on trajectory x,y,z with force n, and that grenades explode in time t, so just take care of it itself. In fact shouldn''t all geometry and objects that are not Player avatars be tracked by each client and rendered based on their behaviors? Then perhaps the server only cares about updating itself and the other clients when a client modifies itself, another object, or creates a new object? I''m starting to ramble...maybe some insight? The Tyr project is here.

Quake1 was developed just like that (and maybe the other quake''s too). it''s a client/server architecture even when playing single player, but the local machine is the server. Really good when making a single player game which also support multi-player.