I am currently using Farseer 2D physics engine on my server code to keep track of bodies in the zone. The issue I can foresee is performance though since I plan on having 100's of these zones, each with their own physics engine. But I am wondering what would be the draw back of not using a full up physics simulation and just use some simple algorithms for movement and collision detection. This game is going to be stat based in the sense that you target something and your attacks / damage are centered around some stat (str, agil etc) and not actually relying on an "arrow" actually hitting the other entity on the screen. I was thinking about using something similar to this and just removing the gravity portion.
Most RPG MMOs do it that way -- run a simplistic movement model, rather than full physics simulation, on the server. Whereas 100 separate physics worlds are too much for a server (or a server cluster?) depends on how many players you have, how fast your servers are, and how expensive the actual simulation is. Simulating nothing (when no players are playing) should be very cheap :-) Also, if you have 100 worlds, each with hundreds of players, then you could spread those worlds across 100 different servers, at which point each world only serves a few hundred players. That would also work, assuming your game business model gets enough revenue to pay for the servers.
Thanks for the reply hplus0603, I was hoping for this confirmation. I think the plan moving forward will be to replace the physics engine and in it's stead have some rudimentary collision detection for the walls, obstacles and possibly players/npc's.
