[lolinder]

It looks to me like "server meshing" is a concept coined by SC for the architecture they have planned. The top page of Google for the term turns up only SC-related results.

EVE uses a different, simpler model where each system is a process, and players connect to proxy servers that keep track of which system they're on (and therefore which server they need to connect to) [0]. I'm sure there are layers of caching and other optimizations built on top of that basic structure, but that's just Gall's Law in action: a working complex system invariably evolved from a working simple system.

I don't think it's a coincidence that the architecture that I can describe in a single sentence is the one powering the legendary 20-year-old game, while no one seems to be quite able to explain what "server meshing" will look like except that it seems to have a lot of moving parts and still hasn't been finished.

[0] https://www.engadget.com/2008-09-28-eve-evolved-eve-onlines-...

[hwillis]

Star Citizen is trying to handle physics for objects in stations/planets/ships, debris, asteroids etc. A couple orders of magnitude more objects, I think.

Note that I don't play or follow either game, but I think you're making an unreasonable comparison.

[lolinder]

I'm only referencing EVE because OP did, but I don't think the comparison is unreasonable: using a simpler architecture doesn't mean you don't use beefy machines and write your code in C instead of Stackless Python to get better performance in your physics simulations. It just means that instead of being overly clever in your network infrastructure from the get go, you pick something simple that works and iterate from there.

[mrguyorama]

It probably starts at picking a game engine that doesn't absolutely suck dick at any multiplayer workload.

SC was from the start supposed to be a giant multiplayer world, yet they picked cryengine, which can barely handle simple shooter multiplayer gameplay. Half of the updates Chris has given on development are just over-explaining simple multiplayer game engine functionality that they've had to implement from the ground up, while you can download unity or unreal and replicate it in the next hour.

I've always found him to be full of bullshit. The kickstarter video had him spend like ten minutes waffling on about how the ships have thrusters on them and all the physics are calculated off those actual thrusters, as if 1) that's not just a useless implementation detail that changes nothing about the actual game, 2) as if that's impressive in 2012 instead of the exact thing the rest of the industry was doing, 3) as if that was hard work instead of just a normal thing a video game developer should be comfortable building, 4) AS IF IT MAKES THE GAME MORE FUN

I don't understand how he sold anyone off that video. The sales pitch wasn't even aspirational, it was clearly and obviously a grift, though maybe unintentionally, from the very beginning.

It's a great example of someone thinking "more simulation detail === more better". Every game developer should really attempt to take a class on the human psychology of games.

[hwillis]

That doesn't work. The biggest battle ever in EVE online[1] had 2,670 players in a single system. The server is chugging through movement/actions/collisions for a few thousand objects, almost all of them not touching and just doing raycasts.

Star Citizen seems to have 100k+ entities per system[2], some with extremely complex collision (eg player models, ship interiors), many of them interacting with each other on each timestep, probably very few of them sleeping.

You can't fix that with faster processors. All those updates trying to go out to players, all subject to checks, all needing updates. It's not in the same class of difficulty.

[1]: https://en.wikipedia.org/wiki/Battle_of_B-R5RB [2]: https://www.reddit.com/r/starcitizen/comments/qbqv7v/heaps_m...