Game Discussions Star Citizen Discussion Thread v12

And it's certainly more complex than WOW or Warframe.
We're not talking about a released final version of server meshing. We're talking about the early initial stage of getting two servers to communicate their state! Dual Universe planned their meshing architecture from the start.
 
they're not talking about anything complex or any large data transfer or -processing issues — its literally about two servers talking to each other).
They just had said they have time synchronization difficulties between 2 servers, nothing more. Me and you know absolutly nothing about the amount of data sent or received, the amount of queries during this test or the number of services running on those servers. I work in databases, when you have troubles with timesync, it's never on simple systems.
Predefined animations are fine, but that's completely against the design ethos of the company.
Where have you heard that ? Server side ragdoll is not a real choice for CIG and between client side ragdoll and predefined animations, the second suits best what CIG want to do.
Client-side ragdolls is also not a problem since the server could still be authoritative on body placement.
Authoritative ? How and when ? The ragdoll occur and when nothing move anymore the server send a new position to the body ? Or the server send the ragdoll position each tick (=server side ragdoll) ?
in a game that has previously given strong hints that it is not actually server-authoritative at all
It's not strong hints. A dev had explained last year that the server is atm client-authoritative and they work to pass it server-authoritative. Normally, it's part of the "Server to Client Actor Networking Rework" card for the 3.11.
 
I'm not sure CR would be willing to give up that amount of control.
Imagine if the player chose to look away and missed his masterfully directed t-posing.

The only way SC will ever see the light of day is if an when for what ever reason, CR steps down or ... And a large company picks it up, and wraps up the loose ends. Any and all new content can and would be in the form of patches and or updates. CR impossible dream is going to be it's eventual downfall unless as I stated, he for what ever reason isn't in charge anymore. In several interviews of previous employee's they all stated that there reason for leaving is CR constant and ever presence of micromanagement.
 
They just had said they have time synchronization difficulties between 2 servers, nothing more.
Exactly. So don't inflate this very basic issue into something it is not.
They say they're “looking at getting two servers talking to each other.”
Don't bring any actually hard networking problems into this because that's not what CI¬G claim they're struggling with.

Where have you heard that ?
Have you ever listed to what Chris Roberts actually says? Have you read the sales pitch? Watched any of their bragging videos going on and on about fidelity, and “physicalisation” and world interaction and reactive and dynamic modelling?

Authoritative ? How and when ?
“Here's the force acting on the body; here's the path it will take.” Same as with every other physics object in the game.
Make it a deterministic system; lock in the parameters (given by the server); let the client sort out the irrelevant animation.

It's not strong hints. A dev had explained last year that the server is atm client-authoritative
True, that is not a “strong hint” — it's outright confirmation. But we knew that all along since, long before CI¬G were able to dig up the dictionary and read up on the meaning of the word, there were strong hints that the server was just a waystation, not something that was particularly involved with the game state or checking up on what clients were doing. After all, that's the flaw that was used by all the hacks that emerged three seconds after the players got their hands on it…

e: Coincidentally, this is why you don't skip off to the mocap studio while another team works on ragdoll animations while a third time works on parametric animations while a fourth team couples all kinds of interactions to the positioning of 3D models while a fifth team figures out a way to decouple interactions from the positioning of 3D models, all long before you have any idea how to maintain a consistent state between clients.

And of course, true to form, this is a long solved problem. CI¬G just don't know about it yet.
 
Last edited:
Have you ever listed to what Chris Roberts actually says? Have you read the sales pitch? Watched any of their bragging videos going on and on about fidelity, and “physicalisation” and world interaction and reactive and dynamic modelling?
Exactly. And the “physicalisation” alone explains why a client-side ragdoll might be a bad choice for CIG. Also the fact that a body should stay waaaay longer in place than in a standard FPS, making each difference in it's rendering more visible for everybody. And also the fact that a body can already be interacted by anybody with body dragging. CIG have no choice, the server must know the exact place and position of a body at the end of each animation.
 
They just had said they have time synchronization difficulties between 2 servers, nothing more. Me and you know absolutly nothing about the amount of data sent or received, the amount of queries during this test or the number of services running on those servers. I work in databases, when you have troubles with timesync, it's never on simple systems.
Sorry for non technical readers, but this is called...
There are really good solutions on the market (like Kafka, Redis..) that allow distribution of data / commands across a cloud of servers, with really low latency (i wont get into the zeroMQ clones or Netty, just keep in mind these ultra low level message brokers exist too), and incredible bandwidth. Redis is basically what CiG call "iCache" but the difference is it's working, and you can run it in like 15 minutes on one node, granted configuring it for a distributed cloud is a bit more involved, but we are talking hours, not months even less years lol. Setting these up on AWS (which CiG are supposedly using..) is a few mouse clicks away.
So yeah if you use for example Redis (which may be the preferred solution for a C++ program like a game), you still have to design the data formats (have to be clever to save bandwidth and such... it's quite a powerful tool and allows everything down to bitwise operations..) and then agree upon which data packets / streams you want to exchange, but again, a working solution will take a few days or weeks at worst. Been there, done that. Yes you'd need the pro license for the full distributed cloud version, but the price is laughable for CiG with their $300+ million.
Mind that quite a few big companies out there do have servers that exchange much more data than CiG servers do, and sometimes in a very sensitive context that allows zero loss of anything, with incredible security constraints (e.g.: finance), and i mean millions of messages per second with no tolerance for any added latency. And yeah we can take a quick estimate at the amount of data that needs to be exchanged through analysis, that would need a few hours of work at most. I understand that game devs would be unfamiliar with these technologies (I kinda see where they are coming from..) but it's really not so difficult to try and recruit specialists.
Disclaimer: this is my day job.

What CiG are doing here, is reinventing the wheel, but so far the said wheel is octagonal with duct taped struts, and they are visibly struggling.
 
Last edited:
View attachment 186499


Edit: Are they going to need a MEGA ship sale to compensate for this ?

Just makes you think that in another year we will probably still be here laughing at CIG's antics while CIG are still talking about making ladders while climbing them.
 
The only way SC will ever see the light of day is if an when for what ever reason, CR steps down or ... And a large company picks it up, and wraps up the loose ends. Any and all new content can and would be in the form of patches and or updates. CR impossible dream is going to be it's eventual downfall unless as I stated, he for what ever reason isn't in charge anymore. In several interviews of previous employee's they all stated that there reason for leaving is CR constant and ever presence of micromanagement.
All current evidence points at CR being put aside. We'll see what transpires at next CitizenCon (edit: oops, ninja'ed by CiG themselves, haha. Looks like we wont know until 2021). If that's the case it's indeed a good news, a bit too late maybe.
 
Last edited:
My goal was not say that SC is as complex as SKA telescopes but that we know almost nothing about the structure and flow of data projected by CIG. And it's certainly more complex than WOW or Warframe.


Thanks to confirm ragdoll server side is stupid. And as Ragdoll client side is not a good technic to have a coherent body placement for everyone, predefined animations can be real choice. Ragdoll client side and predefined animations have pro and cons.


?
Not putting everything on the roadmap doesn't make it a lie. The actual roadmap is short/mid term cards with approx ETA. No ETA = no card.

You're not wrong about the networking and chatting to databases is difficult. I'm pretty sure most companies would look at the potential challenges regarding this and decide whether they can achieve it first and plan how to achieve it before making people give them hundreds of millions to achieve it. Most wouldn't still be trying to achieve it 8 years later either. They would have either admitted defeat much earlier and scaled back what they were trying to do or gave up altogether.

But as long as there are cheerleaders saying "take as long as you want, here's another tens of millions" then that is exactly what CIG will do. Which isn't conductive to getting a releasable product this side of armageddon.

On the up side, it might be the investors are now pulling the strings, and maybe, just maybe, CIG are being pushed to get stuff done.
 
Exactly. And the “physicalisation” alone explains why a client-side ragdoll might be a bad choice for CIG.
…except they've already done it. So they must already solve the problem. So there's no need to use incompetently and inaccurately applied canned animations.

Also the fact that a body should stay waaaay longer in place than in a standard FPS
There is nothing to suggest this.
And even if that was a fact, that would just mean that they'd have to solve the exact same problem, with the exact same consequences.

And also the fact that a body can already be interacted by anybody with body dragging.
Makes no difference.
In fact, that just further highlights that they must solve the problem so there's no need to incompetently and inaccurately apply canned animations.
 
Last edited:
Sorry for non technical readers, but this is called...
There are really good solutions on the market (like Kafka, Redis..) that allow distribution of data / commands across a cloud of servers, with really low latency (i wont get into the zeroMQ clones or Netty, just keep in mind these ultra low level message brokers exist too), and incredible bandwidth.
And you know that CIG is not using Redis or Kafka ? Because for me, I don't know... Icache is just the name of the whole system, not of its ingredients.
And why rag doll server side is not more common in multi games if it's so easy and light weight to implement ? Why GTA5 or RDR2 are not using it ?
 
And you know that CIG is not using Redis or Kafka ? Because for me, I don't know... Icache is just the name of the whole system, not of its ingredients.
And why rag doll server side is not more common in multi games if it's so easy and light weight to implement ? Why GTA5 or RDR2 are not using it ?
I don't know. I have an idea, but it was you defending server-side animations just one page ago. Did you already forget that?
 
Back
Top Bottom