Editor's note: This paper was originally published in the 1999 Game Developer's Conference proceedings.
When we started the X-Wing vs. TIE Fighter project, our goal was to create the first multi-player space combat simulator to be playable over the Internet. There were several major problems that we had to be overcome to accomplish this goal, not the least of which was the Internet itself. I will review the problems we faced, the approach we took, and the results we achieved. I hope the lessons I learned will prove to be valuable to those who read this paper.
The Problems We Knew About
X-Wing vs. TIE Fighter is the third game in the Star Wars space combat simulator series. The Internet was definitely not one of the things that we were thinking about when we created the engine for the original X-Wing game. This was the first problem we faced. Adding Internet capability to an existing engine is significantly more difficult when the engine was not designed with the Internet in mind.
Our second problem was the complexity of the game design. We had always felt that one of the strongest features of our engine was its ability to simulate fairly complex missions. We were proud to have fairly large numbers of craft in each mission, which had reasonably complex behaviors. Our goal in creating X-Wing vs. TIE Fighter was to create a multi-payer game that had this same level of complexity. We wanted to give gamers a multi-player experience that was more complex than "deathmatch." This requirement dramatically increases the amount of data that the players need to have in order to play the game.
Third on our list of problems was that we would not have a dedicated server available; we would have to use a peer-to-peer network model. The expense of providing servers with sufficient processing power and bandwidth for our expected audience size was considered unreasonably high. And because of the nature of the license we were working with, allowing gamers to set up their own servers was not a viable alternative. A peer-to-peer system avoids the problem, but it poses a significantly more challenging engineering problem, because each player must communicate with several other players, instead of with a single server. Because the Internet does not have a viable multi-casting capability, sending the same message to three destinations requires three times as much bandwidth as sending it to a single destination.
The fourth problem, of course, was the Internet itself. When we started the project we assumed that we would need to handle latency that varied from 200ms to a full second. We also knew that we would be limited to the bandwidth available from a 28K modem. These two constraints were our primary focus when we designed our network model, but they would turn out to be among the easiest problems to solve.