[whizzter]

Giving state isn't the problem, the client being authoritive is the problem.

What I'm describing is so-called rollback based netcode, the server should always keep an canonical state and deviating from it is at a clients peril.

[raizer88]

Giving full state is still a big problem. Aimbot, ESP, wall hacking are possibile only due to giving full state to the client. Most competitive games avoid this as much as possibile, eg. Dota2, Valorant, etc.

[whizzter]

Right, I agree there but from a development perspective it's a problem that can/should be tackled later if you know you have a hit, and the Ambient design should have no issue in fixing it later.

- Aimbots is a problem regardless of state it, as long as you can see/shoot at it then aimbots can exist, stopping that is a separate issue.

- On ESP/Wall hacking you're right, it can be an issue

  - Full snapshot/heap shipping as emulators and Kettlewells WASM sample is hard to patch here.

  - an DoD ECS grounded design like Ambient could selectively withhold entities (ie out of sight), however I think for starting development it shouldn't be an initial worry as long as you don't design yourself into a corner in such a way that it's impossible to implement later.

[raizer88]

At the theoretical level I agree with you about creating an MVP first and then solve these kinds of problems the moment you hit the jackpot at the ROI level. The problem is that in 20 years I have never seen a single game do that. Because revisiting the whole network/data stack in a game that has come of age is a nightmare and the PM probably doesn't see value in it. By now you are a hit. The problem is that instead of milking the cow for 10 years because you have a cheat proof product, you make money for 3-4 and then you close. (possibily damaging your brand as a cheater infested game)

[whizzter]

Right, and that's why I like this approach (that I'm doing and it seems the Ambient lib is built for from the examples I looked at) because there is an actual architecture that is designed for networking from day one and visiting netcode isn't as bad an issue because it's mostly just regular game-code as long as the core principles are followed.

We've had generations of games that was built as webs-of-objects (and popular engines that still default to it), and yes, adding and maintaining netcode with that is like pulling tooth because you need to manually keep track of so much.

This puts netcode mostly as an automatable task with clear default rules, stuff like hiding players/entities out of sight becomes filters (in the same way as not updating far-away stuff).

[dgb23]

Only a couple of years ago I played an incredibly fun game with an open multiplayer aspect to it. You didn’t really know who was friend or foe. Alliances and rivalries formed ad-hoc and betrayal was always around tge corner.

It was very well made but had a couple of power balance issues, which is it bo expected.

The problem was that it didn’t get its network code right from the very start. After a fun honeymoon period that was way too short, we would see cheaters all over the place. Not even “just” wallhacking but also teleporting around and dramatically increasing gun RPM.

This sucked out the whole fun in just a few days. My friends and I never played again even though they might have fixed it down the line.