Any idea when the 'base game' is going to work, before moving on to the other stuff?
The Star Citizen Thread v9
- Thread starter Yaffle
- Start date
Well, what I've seen was pretty impressive. I like the way they simulate it, at least voxel-based simulation idea was pretty clever.
To be honest, I really like the way flight-model works in ED but it might be a bit boring sometimes because of dog-fight mechanic. So CIG's attempt to make it more diverse might be interesting. We'll see how it will work out in the end.
It is this far into the project, and especially with such critical core components that need to be finished before you can actually build anything on top of them.
CryEngine's network components is famous for being absolutely horrid in almost every way imaginable, and almost entirely unsuited for anything other than the most simplistic early-FPS style unfiltered, host+client deathmatch gameplay.
Even more bad news: the flight model isn't simulating anything. It is just a massively roundabout way to ensure that a semi-sensible, but very simplistic, physics model is being completely counteracted and over-written so as to to behave in the least realistic or simulated way imaginable. Also, voxels — being a relative, continuous geometric representation of space — have no place in a flight model.
Last edited:
Such project would be a hell for any senior management and I could imagine it couldn't pass the FPP editorial gate in Ubisoft, for example. So it's a, well, miracle they still deliver builds and keep going while expanding the context.
I believe about 80% of their work is tools development and debugging, and I really like what they've showed in CitizenCon - tools are pretty advanced, it really looks like they could create a planet in few weeks.
So we will start to get the 'base game' when they will be happy with their tools, probably.
P.S. I could make an example for Crysis (well, it's not the crown of game design but still) where 3 months before the gold it looked like a cool tech-demo, with everything from UI to basic gameplay elements or AI being buggy or not functioning at all. Suddenly, it become a cool game just weeks before the release when everybody commit the stuff they did.
Server mesh isn't really critical stuff as I see it now. From client's perspective nothing will change, you could freely choose between two server models and still develop.
I couldn't agree, I've seen few MMO's on top of (heavily modified) this network model. It is pretty universal, just because it's all about relying on network serialization mechanic to do the heavy lifting. As a gameplay programmer I just declare the variable and push something into it. What might be tricky to do is to properly mark those variables (there was pretty large choice of possible options last time I seen this). People tend to make mistakes in those. Probably this is what you mean to be 'absolutely horrid'.
They use voxels to model the coarse surface of the ship. And I don't mind if you elaborate on 'isn't simulating anything' - last time I've seen this stuff it isn't looked scripted. =)
Sadly, people have been saying much the same thing about tools for years. I see no reason to assume that it is any more true now than it was then.
It's critical for the kind of silly promises CIG have made. That's the whole problem: they've said a game will do a number of things, but seven years in and they have yet to demonstrate the slightest hint of being able to do any of it because they've focused so much on irrelevancies that they've forgotten to check if they can actually make the game they've sold.
No. You've seen a few MMOs (that alone should be a hint) that have had to completely replace it with something that actually works.
That is why CIG is having all these problems: because they, too, have to replace what's built into the engine with something that is bespoke to their needs and fit for purpose. What they need isn't actually anything special, but it is also not something that CryEngine provides in the slightest. Well… except for the parts where they've declared goals that would require a slight rewrite of the laws of physics. Speaking of which…
They're not simulating anything because their flight model is actively subverting and circumventing the physics engine. They have fixed, pre-determined outcomes, and the flight model only exists to coax the underlying simulation into making that happen, to the point where it rewrites the laws of physics on the fly to create that outcome. It is not a simulation because it is not physics that determine the outcome; it's the outcome that determines the physics.
Also, don't confuse what they say for what they're actually doing. Point clouds have their uses for semi-fast fluid dynamics simulations if you're doing conceptual industrial design — they are not a good choice (and voxels far less so) if you want fast-paced real-time gaming. Doubly so if realism isn't all that crucial to what you want to achieve.
Pretty much any time you see them roll out a
This is the primary cause of their inability to actually deliver anything. Unless and until they stop doing that, the whole project will be an incomplete — indeed incompletable — mess.
Last edited:
What a nonsense....You probably comparing other MMO games done in CE engine that are NOT FAST PACED GAMES that can handle maybe 30+ players per insance but there is 0 games out there that can handle HUGE ammount(100+) of players in same instance that are FPS or similarly fast paced games that are not icon click based games(AKA EVE) where certain lag can be tolerated...ZERO....And this is not only for CE atm there is 0 games in any other game engine that can handle this...And yeah don´t even start with the Planet Side as we allready talk about this before how did they "cheat"to get their masssive battles.....
Last edited:
Which sounds great compared to today's rate but extrapolate and it becomes a nightmare.
100 systems with 3 planets per and each planet taking a few weeks (let's call that a month per planet) and it's now 25 years.
Even a planet a fortnight which would be amazing is still 12.5 years and that's just the planets.
It isn't uncommon to have few branches with something working but in prototype quality and no one cares because everybody focused on something else. So visible to us net-code state might be something really old because its "just works". I have few branches with really advanced AI GAN stuff over our production AI implementation but they are not there yet just because they are experimental. It would be a bit of offence if someone will tell me my deep learning architecture is outdated just on what they see in my 'master'.
CryEngine's network concept is pretty powerful (I mean SerializedVariables, quantinization and adaptive update stuff), guys who written it actually work for CIG now, so it's simply not the case they cannot support the game concept they promised on network level. Micro-services based server mesh isn't something really hard to do, what is really hard to do is to support it (means updating the code after you change something drastically in the game). All kind of hard-to-find bugs might be introduced and this is a real pain because debugging networking in game in just pain in .
May be, and I emphasize may be, it is a wise strategy to have base game working before moving to server mesh implementation. Sometimes you need something simple that can be easily iterated over before building a much more complicated stuff. If I was responsible for net-code development I'd just keep two branches with a straight single-server implementation which I could easily host in my zoo of VMs to develop and debug stuff, and separate multi-server implementation which is just an extension and used only for deploy.
See I'd always go with one branch instead of two - you know, using the less is more approach.
Unless I was going with the - more is better - philosophy in which case I'd go with three or even four branches, let's face it - sometimes you just can't have enough branches!
Either way the whole thing is all so simple - I don't get why it's taking them so long - as you say this is just common stuff!
And never let anyone tell you your deep learning acrchitecture is outdated - that's just rude?
High five!
I wonder what happened to their procedural generation? All the latest stuff I’ve seen has been about carefully handcrafting areas using noise-generated asset palettes. It’s quite pretty though!
I really like the idea of noise-generated asset palettes.
I'm visualising CR sitting close to a mic'd up PC farting to create a carefully handcrafted area.
(I mean it's a lot easier than pecking at the keyboard with 2 fingers - amirite?)
ETA - obv I'm assuming there's no follow through here - that would clearly be the opposite of "pretty" and could in fact create an actual health hazard for anyone else who may happen to use the PC.
Procedural generation is done....superior to everything currently on the market and a crowning achievement. With that being a done deal all thats left is putting in handcrafted stuff before release....what...are you living under a rock?
It isn't uncommon to have few branches with something working but in prototype quality and no one cares because everybody focused on something else. So visible to us net-code state might be something really old because its "just works". I have few branches with really advanced AI GAN stuff over our production AI implementation but they are not there yet just because they are experimental. It would be a bit of offence if someone will tell me my deep learning architecture is outdated just on what they see in my 'master'.
CryEngine's network concept is pretty powerful (I mean SerializedVariables, quantinization and adaptive update stuff), guys who written it actually work for CIG now, so it's simply not the case they cannot support the game concept they promised on network level. Micro-services based server mesh isn't something really hard to do, what is really hard to do is to support it (means updating the code after you change something drastically in the game). All kind of hard-to-find bugs might be introduced and this is a real pain because debugging networking in game in just pain in .
May be, and I emphasize may be, it is a wise strategy to have base game working before moving to server mesh implementation. Sometimes you need something simple that can be easily iterated over before building a much more complicated stuff. If I was responsible for net-code development I'd just keep two branches with a straight single-server implementation which I could easily host in my zoo of VMs to develop and debug stuff, and separate multi-server implementation which is just an extension and used only for deploy.
I ask in all seriousness and not for baiting or being facetious: do you know the names of the netcoders? The reason I ask is that for years, even to the point where CR went on record as crying out for network engineers, they have struggled with networking architecture.
Did they move onto other disciplines? You would think after the performance and criticism of SC networking, that it would be a jewel in your CV crown to say "I was the guy who fixed the SC netcode!"
Procedural generation is done....superior to everything currently on the market and a crowning achievement. With that being a done deal all thats left is putting in handcrafted stuff before release....what...are you living under a rock?
It's just in the other branch, that's all. You know, the super secret one that is working perfectly and has all the features they've promised.
How about seven years into the development of an MMO game which will utterly rely on technology and systems such as the netcode and server meshing...
....and not to jave even started serious design work on how such systems will work and mesh together.
That's where we are now. I'm not talking about game mechanics...but the core fundamentals of the game without which it cannot work. The netcode isn't finished and the server code they need, if we believe CIGs devs, hasn't even started yet. Indeed, parts of it haven't been designed yet, at least beyond the bullet point stage.
Woah, fella!!!!! Them's fighting words. Why everybuddy knows CIG crafted them thar systems uniquely for Star Citizen so claim jumpin' ain't rightly appreciated round these parts
Considering quite a lot of stuff such as server meshing has yet to be added, the idea that hard-to-find bugs might be introduced when it is is a problem that is uppermost in my mind.
Build a game to support 1000 player instances and then find the servers can only support 50.
That would not be an ideal turn of events. You need some idea of the limits of your engine. Fancy coding can help push past those limits...to a degree...but ultimately, an engine can only be pushed so far before it falls over.
CIG would need some idea of how many people its infrastructure can support because that then ties into aspects such as expected traffic load and bandwidth requirements, server resources and player system requirements.
Even if you just used a single server for testing, and remember CIG plan to use servers to support individual geographic regions, you still need a design for the infrastructure to test all those systems work.
This isn't a single player game with multiplayer tacked on...this is an MMO scaled over an entire solar system (s) controlled/hosted/monitored by a number of servers each hosting a separate area/instance/number of players and which need to continually talk to each other in order to ensure the player universe works.
And CIG still say they are designing large parts of it. Not coding....designing.
They'll just need to flick a switch to enable it from amongst the petabytes of data that CIG produce daily - just as soon as that fibre reaches the rack!
Still, the question of comparing old moons to new ones is an interesting one. If their generation tech is actually procedural, then there will be very little need to re-create the older assets as the altered procedure will just kick them off along it's new algorithm and some elementary sanity checking, and of course the showcase handcrafted assets gently nudged to the artists taste.
Now, if the whole thing has to be redone, then either the sanity checks were not sufficient, there is no procedural tech, or what there is of it is overridden by handcrafted asset placement.
Jacques d'Orleans
Banned
Even if they could build a whole Star System per month (regardless of how many planets actually in them), it still would take them 100 month, that's 8 years and 4 month, to do so.