Sony, help us help you!

[Traak]

Remember the SETI distributed computing project? Where vast computations were "delegated" over the internet to computer users all over the world who volunteered their computers for use by the program to run in the background to number-crunch while they were idle of any other uses?

Wonder if we can volunteer to have our computers used as numbers-crunchers for Planetside stuff, to expedite processing of Auraxis terrain, or whatever, encrypted as necessary to keep the game from leaking all over a bit at a time.

I now, it's a longshot, but, they have to have racks of servers and SOME lag that is due to not having an infinite budget therefor, so how about it?

[RedKnights]

Usually distributed computing is only necessary for projects with vast swathes of data. Developing a game doesn't take near as much as say sorting through 100's of terrabytes of audio data or gene mapping. Not even close. It wouldn't be worth their time to bother doing that kind of thing for Planetside 2.

Stop being silly.

And Sony, if you do need help, there's always kickstarter

[Crator]

Not to mention, I don't think that sort of thing is applicable for any type of online game. We don't need to sit and crunch data in real-time for the SETI project and then send that data to other computers over the Internet in real-time. In an online game you do.

[NapalmEnima]

Speaking of distributed computing... IDEA.

The big reason to do client-side hit detection is because it isn't reasonable to have that much processing power. 100 dudes with machine guns opening up on each other leads to a whole crap-ton of math to do to see who did and didn't get hit by what. So they distribute that load to the individual client machines.

But individual machines can lie, leading to all manner of ass-hat-hackery. So don't take an individual machine's word for it.

Every once in a while, client A will ask clients B and C (and D? And so on, at least one of which should be from another empire) if that hit was plausible. Then all the "second opinion" machines report back to SOE... either with a simple pass/fail, or with the result of some section of the heavy lifting that would normally be required in such cases.

If the server sees that all the second opinions are coming from the same (potentially hacked) clients, Red Flag.

If a given client isn't returning some baseline number of distributed verifications-to-shots fired, Red Flag.

If those verifications are coming up negative too often, Red Flag.

Actually, given the horsepower of modern machines, they just might be able to ditch client-side hit detection entirely. Let the clients "ball-park" hits with a relatively bulky convex hull (or hell, a radius/coordinate-aligned rectangle test), and let that flag the server to scrutinize those shots in more detail.

Programmable GPUs are basically massively parallel floating point processors. I'm reasonably sure you can write a shader to do your collision detection for you. A couple extra heavy metal boxen with wicked-sick GPUs could do most if-not-all the number crunching.

Okay, so having 3 machines instead of 1 for each server could get Very Expensive. Fine. Have a pool of number crunching boxen that could float from one server/continent to another, and evaluate everything there for a minute or so, then go on to the next.

Key: DO NOT tell the clients when their hits are being taken for granted or scrutinized. Observe, log, and report suspicious activity. DO NOT let hackers know they're busted until it's too late.

Mwa ha ha!

[Sirisian]

Originally Posted by NapalmEnima

The big reason to do client-side hit detection is because it isn't reasonable to have that much processing power.

Common misconception. The calculations are fairly trivial with proper spatial partitioning (which is trivial to implement). The client-side hit detection is merely for player experience. People don't like shooting at the enemy and not seeing the bullets hit. (This is very important for head-shots). Client-side hit detection is mostly so that people with a high latency can play the game comfortably and shoot at what they see instead of shooting then having the server run an authoritative step and going "no bullets hit anything" because the server state is slightly different than a lagging client state with say 150 ms latency.

I will point out that while a lot of network game programmers like P2P it has a major problem that I absolutely don't like. A game should never reveal the IPs of other players. It's far to easy to DOS attack a single user to make their connection jitter.

[Traak]

Game-included randomly-spaced null shots that are incapable of doing any damage would be one way to catch cheating CSHD abusers. If you managed to score damage or a kill with the electronic equivalent of a gentle puff of air, then you can be flagged for further examination.

[Crator]

Originally Posted by Sirisian

I will point out that while a lot of network game programmers like P2P it has a major problem that I absolutely don't like. A game should never reveal the IPs of other players. It's far to easy to DOS attack a single user to make their connection jitter.

Are you saying P2P would work for a MMOFPS? I would assume it wouldn't but I'm not a programmer so couldn't tell ya. Anyways, I would think the issue with revealing the IPs of other players could be fixed by passing them through a proxy first but I guess that may defeat the purpose of the P2P system then.