#flipcode

Terra Nova is carrying some intruiging musings on the consequences of the absence or presence of player-player collision detection in MMOs. I had always considered the absence of collision detection in WoW to be a technical limitation but perhaps it\'s been left out to avoid griefing.

There is no technical limitation involved with WoW not having collision detection between player models. The engine is perfectly capable of handling that. Nobody would do exact per-polygon collision between player models anyway, so I\'m not sure where the idea that it would be too performance costly to do it (and even if you did do per-poly collision, WoW models are not exactly what you would consider highly detailed meshes to begin with).

The reason it\'s not done is purely game mechanics. EverQuest, the last great game before WoW, had collision detection between player models. The result was raids where melee classes couldn\'t get close enough to hit the mobs they were supposed to be attacking because there wasn\'t enough physical room available within the attackable radius of the mob to fit all of the players. EQ had to solve this problem as the game matured, which meant a lot of hacks like increasing the attackable radius for a mob to some absurd value where you can still \"hit\" a creature even though they\'re about 100 feet away from you. They also had to create potions and spells to shrink players and decrease their volume so that you could fit more than 3 ogres in the same room.

World of Warcraft was developed with this in mind, and also with a strong emphasis on raids (something that just evolved into existence in EQ, it was not by design). The design decision was made that player models shouldn\'t interfere with each other for physical space in the game, and that was that. PvP might suffer slightly because of it, but trying to balance PvP is hard enough without having to worry about any particular race having an advantage/disadvantage due to their size.

In my estimation, there are good reasons, technically to trade off player to player (and other dynamic models) collision.

The obvious one of these is latency. Collision between players/human controlled objects is the worst possible latency case and the hardest to handle with prediction. With no player collision, you reduce your player movement/prediction problems a lot - and if you use a mechanics/square turn based combat system...

Of course, all this reduces your bandwidth usage to a trickle, and your high latency rates mean you can cram a server full of players without worrying about turn around time.

One of the things about the WoW network system is that the movement and combat is pretty smooth, even with a very high latency of 400+ ms.

To achieve this, the client renders surrounding players a large period of time behind their \'actual\' position. I\'ve seen this in action with me and a friend playing on the same server, running along a road together: on both our comps, the client shows us to be in front, and the other person 30 feet behind.

So I imagine if there was PvP collision, the displayed latencies would have to be lessened, to something more typical of a FPS. And this in turn would put a bit more load on the server, i guess. Also player movement physics is computed on the client in WoW, so that would make PvP collision kinda tricky :)