Sorry your browser is not supported!

I have made a few online games before but never a 3D one. I would like to use GDK for the client but the server be a normal windows console mode.

The problem is how does the server process 3D movement? or ray-tracing a bullet? etc...

I see 2 choices:
- do it manually, using 3D math that I need to figure out, yuck
- get rid of console mode and use GDK on the server which could cause performance issues and would need odd tweaking to support multi-areas, I think

And I can't put the processesing on the client... NEVER TRUST THE CLIENT!!!

OK, thanks!

Hi Grayvyn,

Unfortunately, letting the server process 3D movement would require a lot of power, especially for several dozens of players simultaneously.

Most online games let the client do the processing and then sends it to the server which checks if it's possible, if it isn't, it would boot the player for cheating.

Hope this helps.

-Mike

Actually they don't... Most online games calculate the movement both on the client and the server, but only the key-presses are sent to the server, so the client calculation is just to make it feel more responsive. Calculating 3D movement is not particularly processor intensive either.

Seeing as the server needs only the collision data, why not use a collision library for C++, and use that instead of GDK.

You have, unfortunately, contradicted yourself. Also, if the server calculates movement of every player, games like World of Warcraft, which has 3000+ players per server, would be unbearably laggy.

And, collision has nothing to do with the server as all the world data is on the client, else it would be ridiculously complex.

Hope this helps.

-Mike

Where?


I can't see that calculating movement for even thousands of players would be very computationally expensive. Well, the collision might be.

I do know for a fact that many first person shooters do as Diggsey mentioned, but things might be done differently in larger, persistent online games.

My client-server system uses a distributed load. The server assigns each client a group of objects to simulate, where the group size is based on how the client's computer rates. It checks the CPU speed and checks if the video card has PhysX hardware acceleration, or if they even have a PPU, then sends the info to the server, so it can assign it. If the client disconnects, the server takes control back. It also does some reasonability checks to make sure there aren't hackers. The server just takes control of anything not simulated. It also won't give the client anything if their computer can't handle any more than a few objects.

This only works for a massive game, though. Like, an entire galaxy. Otherwise, you'd get issues with small groups of objects, like if something actually collides with something from another group. That'd be really troublesome, the client wouldn't know those other objects are there.

It does a lot of other checks when assigning objects. Like, if the object in question is moving fast or is near something moving fast, it'll keep all the included objects to itself. You can always control the server quality, but never the client quality.