#server-plugins-read-only
1 messages Β· Page 31 of 1
Can't really say for sure until we got access. I wonder if world generation is a big one for CPU usage because for any non-SMP servers thats a no brainer to disable
They likely can't either yet.
They are testing for a reason! π - Many more tests hopefully
They had a test with 56 players last week and it crashed because of a bug of the server and they had over 3000 npcs.
Now they did a test with 70 and it crashed after 45 minutes. That's 14 more players
The server will get optimized with time
And more unknowns! π (not a bash) Just being captain obvious. π
Gotta remember this game is like 2 years away from a release
2 years? π Idk bout that one π - Its a bit futher
They said the engine was like 1 year from being ok for release
I think 2 years might be the number they said as the minimum for early access
In the end, we'll have access to the server code so optimization will also come from the community
Yeah, I'm curious how that will end up going
We'll probably be able to make some really flexible tricks for scaling
Engine and Game are different. I would think the Engine has 2 more, the game is longer.
From what Simon said, the game will never stop being worked on. It will probably receive expansions like an MMO would (Simon does love World of Warcraft ater all)
Oh yeah, I like this statement. - I was excited when I read it
I really liked when he talked about cinematics and voice actors too π
Im not much of a fan of that stuff more like just freely getting to do what I want. π - But this is still good news!
The way I see Hytale, it seems to me to be a mix of Minecraft and Skyrim. Sure it's not exactly that, but the combat and ambiance of the game really looks like the same.
Now what's cool is that I really really like Skyrim, but creating content for Skyrim is very hard. Hytale would let it be much easier and I love that.
I think voice actors is interesting, but not something I'd want tbh
I recently joined Wynncraft again for the first time in like 10 years and the voice acting was certainly an interesting addition
Turn off the voices and use subtitles π (Not for Wynncraft but for Hytale)
Is that a thing on Wynncraft?
Wynncraft has voice acting now? WTH?
I find it more immersive when you don't have voice acting, because it's rarely a studio-grade recording/performance
Been a while since I played Wynncraft and even then, I never went far. I'm not much of a fan of MMOs. Especially not when I can't just play solo
But Hypixel Studios Canada IS a studio π
I would def use voice actors if I could
Are you guys pro or against AI voice actors?
I think conveying the voice through text styles is way more interesting
Even if I or someone else doesn't like voice acting, the fact they want to go that way tells me they have a clear direction contrary to Minecraft that stays more "vanilla" or "general".
It is already difficult enough for voice actors to find work lol
We should not be replacing them with lifeless mimics
I agree
AI voice actors for a mod? Sure.
AI voice actors for a studio? Common bruh
Since we are in hytale, I wouldn't be surprised if someone made speech bubbles or a zoom in dialogue type of thing
I thinking having no voice acting in Hytale would make it easier for mods to mimic the style as well
There will probably be a dialogue system mod real quick.
Conveying the voice through text styles is way more interesting to me anyway
Def, we have proximity though, so I guess it depends on your game design. Voice acting + proximity makes it hella immersive ngl
Oh maybe that's a good point
The way I see it, you could have either voice acting or just NPC mumbles with text
β
Someone is going to make a speech to text plugin and have you talk in the mic to continue π
I like mumbles too, or the muffled dialogue some games do for it
I remember Assassin's Creed Odyssey that let you create your own quests/stories and the characters talking would just be mute. Adding a mumbling would have been much better
LOL using player microphones to alert nearby enemies to your presence
Or cause cave collapses maybe
yeah
Use AI to let NPCs react to what you say
I know roleplay servers are going to go crazy with this one
NPCs that give you riddles that you can answer with natural language processing would be interesting
Like getting into a dungeon requires you to answer a few questions
But you could ask for hints or give synonymous answers
With something like that, it probably wouldn't be too intensive because they're one-off tasks, you'd need a lot of players using that particular mechanic at once
A friend of mine was doing this. π - Yeah it was fun!
Plugins and mods are going to be paid or the eula says they must be free?
You can put your plugins / mods for sale yes. (or behind a subscription model [example: patreon])
As long as it follows the EULA and community guidelines
animal crossing
Does anyone know if server plugins will support kotlin?
Also iβve not done a lot of java modding but is there a harmony equivalent for java?
quarkus or spring boot. anyone here ?
will hytale natively support cubic chunks?
I was thinking of doing custom terrain generation for beautiful infinite worlds at some point maybe
and I am hoping that I can combine that with custom vehicles
The chunks are in 32x32x32 (cubic), but the files store them in columns of 10. It's something they already plan on moving from so we can get infinite height worlds.
Hmm looks like mixins are the closest equivalent
do we know something about external oauth integration? such as authorizing on your website by hytale account
@late breach hytale oauth wound be dope
You could probably build a simple server for this
You just need someone to log in-game on a server and type out a message right?
That would be a fun first project. To make an account verifier for hytale.
true
good point, i already have solution for web auth approval, but oauth would be great
Yeah, that's the only safe way currently.
User Command -> Store Hytale Name in a API Server -> Generate Token -> Show Token + Link where to verify -> User enters Token -> API Request -> Verification done
I wonder if there's any reverse proxy that works with QUIC protocol
wait we can't run a hytale server then set a simple nginx reverse domain proxy to it?
i dont know actually
That would be Nginx lol
nginx*.*org/en/docs/quic.html
nginx seems good
another concern is would the backend servers be able to recognize the connecting players IP addresses
I remember that sort of stuff being quite annoying to deal with last time I played around with UDP proxies
Even if they can't, why would they need to?
ip bans / preventing server scrapers
You can do both of those from the proxy itself
maybe sure
that does mean i would need a custom system if, for example, i want my moderators to control the ip filter
but its fine
You'd need a custom system regardless if you intended the bans to work across all of your servers lol
You just read the IPs from the db, just like every other server in your network
maybe i dont need to retain original ip after all
the only reason i need a reverse proxy is because the actual host (which is my friend's machine) can't port forward due to isp limitations
so i have to put a secondary machine on the cloud and use that as the gateway for players
oh
if nginx can handle hytale's udp traffic well it should be all good
Idk how Nginx works in this situation, or how they have it implemented in Hytale
But it's not entirely impossible that there will be a way to recognize headers like X-Real-IP, which would give you the original
If there is, then you could just use normal IP bans
that would be convenient
For like a bungeecord network?
Afaik hytale devs said that stuff is already built in
cant wait to cook
No, he's trying to get around ISP limitations
Hytale does have a transfer packet
all i could find was a transfer packet and not a reverse proxy
transfer packet just makes the client connect to a different server
Could ask slikey tbh
There are like a dozen reverse proxies for QUIC
You don't need to ask anyone π
Can someone tell me in mc what exactly is a reverse proxy?
It's not a Minecraft concept
I know, but an mc example if someone could give
It's literally not a Minecraft concept, you can't just give a Minecraft example lol
ah
users connect to one server and the connection gets routed to many backend servers (thats just one use case)
ohhhh I see
good reverse proxy softwares has more advanced features like load balancing but ive never had a server large enough to ever need a load balancer lol
Technically every time we use "proxy" when talking about Minecraft servers, they're all reverse proxies
Because the difference between a proxy and a reverse proxy is just who is in charge basically
It's just that on the open-web, we don't just say "proxy"
Because the difference actually matters a lot
i dont think hytale would come with an official reverse proxy system because that'd mean the code has to be extremely flexible enough to allow handling packets without real "game server" behind it
You don't need one, and you wouldn't want one
You can use the industry-standard and fool-proof reverse proxies lol
yea except you dont really get to control stuff based on the in-game context
Like what context?
commands etc.
i mean server can still communicate with a general purpose reverse proxy software via various means but
Why would your commands control your load balancer and handling encryption π
i mean redirecting connection to a specific backend server
βAfaikβ they only said the needed protocol package to transfer players between servers is implement not that it is a ready to use plug and play solution yet.
ok
Can you give an example
basically what /server does in bungeecord/velocity in minecraft
You don't want your reverse proxy to handle that to begin with
That's the game's job, and the game supports exactly that natively
Why not transfer packet?
i feel like transfer packets are easier to spoof by client
WHAT LOL
uhh ok my example isnt great
No offence but why are you worrying about those little details? Why not focus on making a good server first. Like nobody cares about those geek stuff
How is that a concern at all for routing
You'd need game support either way because you can't just drop the connection and move them to somewhere else
That hack works in Minecraft because we tell the client we're entering a new dimension
You have no idea how many times I've seen people create orchestrations, custom systems just for their server to have ZERO players
So rather than using a hack to trick the client with a bunch of fake bs, we just tell them they're transferring directly now
yea i see now
Their concern is a security concern, not "geek stuff"
But it doesn't make sense to me
idk man I just know the basics, when you overlook stuff you are cooked
I am going to jinx it but who's going to client spoof you anyway
honestly i was thinking of the features of velocity/bungeecord thats not reverse proxy so it got a bit off topic
such as communication between backend servers
Hi, do we know the minimum and/or recommended server specifications (pure vanilla) to handle x players ?
We won't need Velocity for transfers at the very least
Communication between servers is probably going to be built-in
You'll be able to share databases between them no matter what, so that much is fine
We don't really have recommended server specs, but they posted some of their testing results
slikey said that bungeecord system is costing them thousands for their minecraft servers so im pretty sure they put a built-in solution to help fix this and this person is concerned about a nothing burger in the end.
I wouldn't call it a blog post. It is more of a technical manual with some example data. Just wrapped up a playtest with 70 people on various hardware until we finally crashed the server.
There is a lot of context such as "minigame server" vs "exploration mode SMP". We believe that a 3 vCore + 8 GB server can sustain 6-8 players. A Ryzen 7950x with 12 GB RAM server, we were able to push 70 players but TPS degraded until the server crashed (~45 minutes).
I think we will defintiely see 200 player minigames such as Mega Walls or Super Sky Wars. While vanilla SMP is going to require some beefy hardware to push 100 players in a single SMP.
We are continuing our effort to improve stability and further add performance optimizations to get those big anarchy servers.
my* minecraft server (not exactly mine) uses redis for backend communication but a standard system for that kind of stuff would be nice instead of leaving it up to each server developer
Ryzen 9 is vague as hell and it failed to "handle it" lol
Honestly I think that setup can def handle 500 players for minigames
We have no idea what kind of scaling issues they have
I am going to test it day 1 ye
Where this quote comes from?
hmm, okay, thanks
Didnt they confirm servers can disable unwated features? and that in return can cause performance boost?
slikey/status/2000608313977810946
yep that's confirmed. The whole game is plugins basically, you can disable whatever you don't want
Yes, but we still don't know what the bottleneck to begin with is
Dude isn't it crazy that you can just drag n drop a new plugin and BAM you have new features
That's kind of what plugins do anyway LMFAO
Yeah but like I'ven ever seen that before
this seems kinda on par with minecraft servers which is pretty good i think
I think that's because you're not a developer π
facts
You think that's on par with Minecraft? Are you crazy
Vanilla Minecraft cannot handle anywhere close to 60 people without TPS drops
wait, a long time ago one of their blogposts said that servers can change textures, is that still up-to-date info or now not the case?
3 vCore + 8 GB server can sustain 6-8 player
at least this part that is
Vanilla Minecraft is so poorly performant that there probably isn't a soul out there running that software for their real server
i dont know about how it scales at large number of players because my smp only ever got about 15 players at the most peak
I hope that all these purchases will enable them to optimize everything properly, because the server is still quite demanding compared to other voxel games
Yeah it can change everything, ui, textures, animations, vfx
All content is data-driven
The client doesn't have anything, so yes
i see, thanks!
20 is the default player limit and it's because it'll have a stroke on most hardware π
And that's day 1 btw, they will introduce optimisations later
One thing about the whole reverse proxy stuff is that if you just use transfer packets, all servers need to have a public facing IPs whereas with reverse proxy, none of your servers need a public IP, the public IP of the reverse proxy is enough
that is true, but isnt minecraft kinda unique in that regards
i dont think a lot of games take that approach of like connecting everything behind one reverse proxy
what about cheating? Is there anything planned?
their EULA states server-side anticheat.
Hytale is apparently going to work like that too if you use nginx for example
not sure if Traefik supports QUIC
traefik for me is for home deployment, not for prod
Okay so with the transfer packet, you have a payload
You can send this player a transfer packet to the reverse proxy itself, and the payload can dictate the route
So you don't need to expose every back-end server, you can just re-use the proxy for the routing
Hm, I hadn't considered that with a transfer packet we need to expose all boxes.
ah thats exactly what i was thinking, an additional payload (which MC does NOT have π)
You don't need to, you can send the player a packet that connects them to the reverse proxy they're already connected to
Basically forcing a re-connection
Or maybe some kind of proxy setup in front of all boxes? Not sure how that would work, but if there is an additional payload a reverse proxy could use that to route correctly
If you want to have a bunch of servers that players can jump between, but cannot connect to directly
You still use a proxy, it's just a typical reverse proxy
Not a specialized Minecraft one that abuses the client
if you have a domain cant you also like do transfer to a.mydomain b.mydomain
and have nginx handle the subdomains
Yes, you could do it like that
But those are publicly exposed boxes
yes, you don't need to if you have a reverse proxy, however if you don't, you're gonna have to have a public facing IP... I don't know how many server providers will also offer you a reverse proxy if you ask for 2 or more servers
Can you send a player to any arbitrary host, or is there configuration needed such that the sending and receiving servers know about each other?
Ah networking... my worst enemy I have not thought about yet. I hope someone in the community will have a proper bungeecord-esque implementation for the Hytale transfer packets π
Maybe, but DNS can be a bit finnicky when you have automatic spin up of boxes and shut down
I don't think I understand what you're saying here
Why would they provide a proxy for you?
I think servers with heavy modded content will need to use a FastDL CDN right?
Yeah, I think so
Idk if they've addressed that concern yet
Not the Hytale team but server providers, if you want to host a Hytale server or two but don't want to exactly fiddle with a linux server
fastdl oh no
havent heard that since gmod days
I am not so familiar with QUIC, but it seems like neither Traefik nor haproxy support it (or at least not a custom protocol). But maybe I am mistaken?
me spending 2 hours downloading a server's 80GB content:
will hytale allow us to customize loading screens like in gmod π€£
If you want to host a Hytale server or two, and you don't want to expose all of them to the web
You just run Nginx or whatever for routing
If you expose all of them, you can just use the transfer packet easily
i think that might be the case tbh but idk
QUIC is like HTTP/3 basically, so there are lots of proxies for it
Including Nginx
gmod allows you to send custom html webpage before the client starts downloading the contents so players have something to look at (and even play music) while downloading
Stupid question then as I don't fully grasp QUIC, does Hytale use HTTP/3 then, for all traffic?
yup, but I'm now talking about the people that want to do that too but don't want to run their own Linux server so they will use some other Hytale specific server provider like Nitrado
HTTP/3 is just like, another thing that uses QUIC i think
i DO hoped for that awhile ago..
QUIC is basically just a modern replacement for TCP, essentially
It's technically possible they don't, I think
But I don't fully grasp QUIC either
No, QUIC is a protocol built-on top of UDP
Because to my understanding HTTP/3 sits on top of QUIC, which sits on top of UDP
yes but like the goal of QUIC is to do the same thing as TCP and improve things
it just uses UDP as the ground layer
Yeah okay that's fair enough
According to GPT, the right stack is:
HTTP/3
QUIC (which is UDP)
IP
So, lots of proxies out there support HTTP/3, but not arbitrary QUIC implementations (not HTTP/3)
It's a bit like why using Traefik for Minecraft proxying would be weird, as Traefik is primarily a HTTP proxy
people decided TCP sucks so they started building a new protocol under UDP
and even the web technology is going to move over to QUIC in the future maybe
We know that Hytale works with Nginx due to a tweet
So although it could technically not be HTTP/3, it sort of seems like it might be
Nginx is one layer lower I think, it can operate on the transport layer. But my networking knowledge is kinda bad at the moment, wwould have to read up on it again
garageagle
Can transfer packets still work if your servers don't all have public IPs? Seems like that would require some sort of proxy
Slikey
Yes, then you need a proxy but you don't need a bungee cord then, you can use a simple QUIC proxy such as Nginx to route your traffic.
I didn't realize they covered this exact matter, but yeah
ive only used nginx for TCP, and specifically HTTP stuff which is the main reason why nginx exists
i dont know how well nginx performs when it goes out of web server things
From what I understand, it's what everyone uses basically
seems promising then
Though looking at it more closely, Jeff might be right about it being a layer lower
It's possible we won't get headers or anything from HTTP3, but we don't really need them, and probably don't want the bloat of HTTP3 lol
i hope we can customize a server's loading screen π
It would be a little unusual because we're not moving text, we're moving actual data
please do not make downloading gigabytes of contents the normal things for servers to do π
that might happen anyway
I'm not sure if that's much of a real concern lol
It'd be nice to show some information while the player loads in for 8 minutes with their 1900MB download though π
custom loadscreen in gmod only exists because servers started putting tons of contents to download
their previous server listing concept allowed us to post information and screenshots about our servers before a user joins...
Gotta put a whole movie on the loading screen because of gigabytes of content
Yeah, but I'd imagine the loading screen would hide that
I suppose it could just do a spinny circle or whatever
I mean the bee movie fits in a GIF, if you can accept slightly reduced quality π
i wish i can post media man
I don't think it's that important whether you can change the loading screen lmao
can you display html elements to players while they are in server
considering it's during the time you download stuff from the server I doubt it will be possible... but maybe when you're transfering from a server to server then maaaaaybe?
don't give the advertising companies ideas
it depends on how customizable the ui would be in hytale but it'd be funny if people start building react like tech
some fancy minecraft servers actually do that in their own propriatery code
Cant wait to play.. literally only 27 days left
Yeah I have seen servers with dynamically loaded ads based on region being displayed that send you to other servers when clicked etc.
Do you think it's worth paying for 1 big server instead of owning multiple small ones?
I don't need to see Hytale + React + TypeScript running together, I've seen on the web and it can stay there
if hytale allows servers to dynamically customize ui elements its inevitable someone builds react translation layer π₯΄
Thankfully the ui is just a java wrapper π
Bad Apple with Hytale UI elements when
3rd day after EA launches... nah, jk, idk
I will be def cooking my own ui wrapper.. html / CSS style
I hope they let us draw a singular triangle of any size π
if Hytale allows a webview component, it's so over... Hytale will then become another web browser
I doubt they're going to give us a web-view component
I wonder if the UIs will be 3D in any way
Like, if you can have an inventory model instead of an inventory texture
Peering into a backpack, with an animation and everything
oof that would be nice
I think thats how it is, considering the whole game is plugins
If they're not 3D, I'd imagine there will be ways to open a dedicated 3D one eventually
Otherwise you just have to put it in the game world I suppose, but it'd be simpler to avoid that
I mean, they did show a 3D player character when you open the inventory right? So it should be possible theoretically?
Yeah if they let us, that kind of thing could be reasonably hard-coded though
And since we're dealing with UIs, now it's a client thing
So we'd need a pretty mature client API
I wouldn't be surprised if this kind of thing is unavailable at launch
I would probably limit myself before I get sugar rush from too much candy. Everything seems possible
what...
Best I can tell from the Creative Mode Gameplay video, it may be locked to just being able to display 3D model as a background and not as a front element
Isn't the creative mode a plugin too?
Slikey said some features are not yet plugins but eventually will be
Machinima panel too
Simon keeps saying server sends mods to clients.
Yeah I'm genuinely curious how far they've gotten with UIs provided by plugins
Damn I wonder if the machinima section is client side or server side
Remember the code snippet for ui?
That was a deprecated system, wasn't it? Like one they ditched later
High chances the creative tools are plugins
Built in afaik
Deprecated in what way? they mentioned they are moving to noesis
That would be an incredibly impressive reference plugin to look at
fr
Studio grade build tools, help UI and everything
Simon's definition of "mods" is way different than a Minecraft mod. Under his definition a Hytale Adventure Map created by community is a mod
Dude, of course and it's a plugin because the ui snippet is in JAVA, the client is in c#
I just thought about it π
2.8 "Mods" means modifications, plugins, scripts, maps, skins, models, and similar content users may create for the Game, which comprise users' original creative work, subject to the terms and conditions in the EULA.
So mods/plugins are a thing. Cool.
This is indeed a definition for a legal document, so it might not be the colloquial definition he intends, but it's one floating around
It's a catch-all to make sure they don't miss anything
The real definition of mods.
someone had the ridiculous idea of making an hytale (partial) implementation of bukkit api π
in theory you could have a really abstract api that works for both minecraft and hytale but it can't do anything specific to either of the games
yeah, but who really wants that lol
probably would be fun for somebody's random weekend project
someone will be dedicated enough... like those people that want to rewrite the entire server in a different programming language
oh yea someone will rewrite hytale server in rust to achieve that β¨ blazingly fast β¨ performance
Rust is probably more about safety rather than speed but faster than Java probably, if done well
rust compiles straight to machine code, so it's in the same ballpark as c++
hytale server in pure assembly
Many things compile to machine code, but machine code is not all the same yk. It's arouuund the same ballpark but also not really
depending on how you look at it
rustc literally uses llvm as a backend, do you even know what you're talking about lmao
you can compare rustc instruction output in compiler explorer with gcc or clang
Yeah clang too, but generate the same code in both languages into LLVM IR and you'll see that it very much depends on usage
Hi, I'm new to Discord, is the server example available yet? I'd like to host it on an Ubuntu VPS, would that work?
sadly not yet
i really wish they provided at least some basic testing tools to ensure the server setup would work when it releases (especially useful for testing UDP/QUIC routing setups)
No, it will be available when the game comes out. Ubuntu VPS will work, you only need to set up Java
They're planning to release a selfhosting guide a short while before the release.
It would be great if they made the server repository available on Github, and provided an easy and simple way to install what's needed to run the server.
I mean Release is not really meant to have Servers already up and running, it wouldn't make sense to put that much effort into releasing tools just for a few days of headstart when everyone will have to explore the game on the first day anyway
Yes the server code will be published some time in the future.
The only thing that's really needed is Java though. pretty simple
ill probably just make sure some basic ping pong connection with a quic server/client works
and hope it'll work when hytale relases
do we have information on what port it uses by default, or a preferred port though
i dont think so
im hoping the server browser makes it so that players don't have to manually specify the port
Did they already mentioned something in regard to offline mode or do you guys expect I would need multiple accounts to test stuff for more players π
Hmm have any devs commented on serving content from CDNs?
I will make a rewrite of Hytale server in C# using unmanaged code to achieve this kind of performance in performance critical scenarios
that honestly sounds just like resource packs in minecraft
with the extra functionality to directly send data to players from server (like minecraft bedrock)
Rust is good for memory safety and performance, but a mess about the plugin system etc. The learning curve is insane to make something simple
With integration of harmonyx?
Dont need to use harmony π
Harmony is used to patch binaries, so its not needed in my case
I will allow devs to make their own plugin and loading it using native C# lib loading
Hmm, guess youβll have a way to monkey patch natively right?
Speaking of, do you know if they build the client using NativeAOT? If so that probably stops you from using harmony lol
Yes NativeAOT is used
i'm planning to write a server implementation in rust and use wasm components for plugins
minecraft community went too crazy that we already have several people planning to rewrite the server for the upcoming game that has not been released yet
I dont expect yall to have feature parity. π That is not a realistic expection unless Hypixel is slow moving. π
Itβs gonna be a like that xkcd panel with 20 competing rewrites lol
Competition is always good! We get choices!!!
True
Havenβt written Java since first year university lol, should probably brush it up
Nah! π
syncthing is an INSANE PROGRAM
ive never used it before... was toying around with Vast AI gpus, was rendering a minecraft animation with 4 gpus
Is the january 13 release expected to have the ability of making working multiplayer servers?
yes, just a program that sync files between two pc or more
yes
aight thanks
And then many of them will realise it's actually pretty hard to do
If we can decompile the server itself, we can now packets
If we dont, wireshark will become my new best friend for few weeks π
Hello, I've started following Hytale news again after the pre-sale began.
I saw some statements from Simon online, mainly about servers; there have been many changes since I stopped following the updates.
Could you tell me what the base language is for developing plugins and mods in Hytale? Also, has there been any statement about "unofficial" player access to the servers?
Thank you in advance.
Java
You can't mod the client, there is no plan to do that
Language for plugins/mods: Java
Clarification: Mods are all server side. The client is not moddable. The client is in C# (compiled with Native AOT).
The game will be accessible to everybody on january 13th.
Additionally, they plan on having node scripts (like blueprints in Unreal Engine) for non-programmers. Models can be made and animated with Blockbench. Assets of the game can be modified with the Assets Editor.
Whats the .NET version used for the client side ?
Used to be .NET Framework 4.7.2, but they moved to .NET 10 since the acquisition. They also got rid of the platform specific dependencies.
All good π
Keep in mind they're using AOT to compile it down to machine code so editing the client won't really be possible
Reverse engineering π
But that will be a ton of work
Does anyone know the server requirements? I want to create one when the game launches.
I wouldn't call it a blog post. It is more of a technical manual with some example data. Just wrapped up a playtest with 70 people on various hardware until we finally crashed the server.
There is a lot of context such as "minigame server" vs "exploration mode SMP". We believe that a 3 vCore + 8 GB server can sustain 6-8 players. A Ryzen 7950x with 12 GB RAM server, we were able to push 70 players but TPS degraded until the server crashed (~45 minutes).
I think we will defintiely see 200 player minigames such as Mega Walls or Super Sky Wars. While vanilla SMP is going to require some beefy hardware to push 100 players in a single SMP.
We are continuing our effort to improve stability and further add performance optimizations to get those big anarchy servers.
That's what he's saying, reverse engineering a native aot app will be much harder
Sure, that's why im saying that will be a ton of work π
I understand, being developed in Java is a bit worrying, at least in relation to Brazil.
I don't have a very in-depth knowledge beyond developing plugins as a hobby in Minecraft, but Java's poor performance in relation to high memory consumption has always been a problem.
Besides, VPS or even shared servers with a reasonable amount of memory here don't usually have a very attractive price.
What is source shared in this project ?
The server code
thank you so much
Good to hear
At the game launch or later ?
Minecraft is made in java. Most of the modding community is using the java edition. It would be dumb to use another language
Later. Last I heard, they said "1-2 months after launch" because of legal stuff
It's unobfuscated, so you can just decompile in the mean time
C# looks like java π
Perfect
Trust me, I'd love for mods to be in C# instead of Java lol
That's my goal too
That's why i will sent weeks to make a C# server
You won't be alone.
People that want to make a server port to C# should try to get together
But way cleaner
It's understandable, but with the ease of access to information nowadays, it wouldn't be so bad to seek knowledge in another language.
Especially if, in exchange for this "work," we get better performance compared to other servers.
Honestly if someone gets kotlin plugins working Iβll probably use that over Java
Java performance isn't bad
The server code will be open-source. The community is already planning on doing ports to C# and Rust
i think y'all are crazy lmfao
it's an early access title and the server can already handle double the number of players a vanilla minecraft server can
because all the content is server-side, you're going to be dealing with new features 24/7
they already confirmed that there are going to be nightly-style releases, and that they only plan to support the last few releases, if not exclusively the latest release
Shouldn't be an issue. A jar is a jar is a jar
NΓ£o Γ© ruim definitivamente, e estΓ‘ bem melhor do que jΓ‘ foi no passado.
Porem deve se admitir que hΓ‘ opΓ§Γ΅es mais viaveis, e com isso fico contente em saber que a comunidade busca criar uma portabilidade do servidor para outras linguagens.
Sure, but C# can be far better in a lot of cases, without speaking about performances
?
? English please
It's definitely not bad, and it's much better than it used to be.
However, it must be admitted that there are more viable options, and with that I'm glad to know that the community is seeking to create a server that can be ported to other languages.
I made a mistake pressing ctrl + c
No, I don't think it must be admitted that there are more viable options (suggestions?). Java is an excellent choice for this project, given the extensive experience of the Hypixel team, the target audience, performance requirements, and the core goal of moddability as a first-class citizen.
also being able to rely on the JVM for all of the low-level stuff (platform compatibility, memory management, tooling, etc) is a huge plus for development velocity
It will be very interesting to see if alternative server implementations will manage to stay relevant if they give up compatibility with the Java-based plugin ecosystem.
(And if people build a shim, if that doesn't eat up all the performance benefits the alternative server would have)
I could see big networks like Hypixel doing their own server implementation with their own custom plugins
i think it'll be possible to achieve plugin development on another language with an alternative server implantation. the real pain is being able to make the api work with it
Itβs also probably the only way it was possible to ship Linux & Mac at launch, it wouldβve been infinitely more difficult to make a stable build for multi platform in early access if it didnβt have Java at the core
with how many people are clowning on java (including me), you really can't forget this quality
I mean this is exactly why wrappers existβ¦ Iβm sure someone will make βgo4taleβ or some dumb name that wraps the hytale api for go compatibility
You really can't imagine how many game servers are Windows only 
Hi-C (c++ wrapper)
finally i can run a backend server on my hytale server 
Java hate is so forced π built in gc/mem management, strong OOP, interop with kotlin, and in 99% of situations it is as efficient as C++ if you know how to write good code and use stuff like project loom and other newer language features
Not really, the client itself is in C# not java, and the hard part of portability are the graphics/audio APIs, which java/C# don't really fix
agreed. It just becomes a meme at this point and people take it seriously
The client is fully C#?
Yeh
Relevant here, new tweet from @random magnet (1/2):
If I had to add one more thing: the modding strategy I posted has flaws. I said that explicitly, multiple times. It is incomplete, and we are continuing to iterate on it.
One concrete example is client-local scripts and client-local asset packs. I did not properly account for the amount of client customization that is possible without client modification. Through our equivalent of data packs, we can actually do quite a lot here.
Yes, this approach appears limiting. Yes, Hytale is more limited than Unreal, Unity, or true Wild West modding.
Those limits are not arbitrary. They exist because they allow modders - and us - to reach a massive audience, keep players safe, and build a UGC ecosystem that can scale. Our goal is to push UGC into a new golden age, with a model that simply hasn't existed before.
On features like shaders, LODs, cubic chunks, etc.: I love these topics. Trust me, we want to make Hytale the golden standard for block game technology. How that all looks years from now is unpredictable. We work on many things we don't talk about, and some of them may eventually materialize in one form or another.
We have very high aspirations for Hytale but for now: We will release ASAP. Take LODs and cubic chunks as an example. We are already working on chunks and voxel storage at a level more fundamental than cubic chunks. This is an order-of-operations problem. One step at a time and the next step is.
So I don't want you to fish for promises - because we wish our promises to remain special and reliable. Judge us by our actions, and hold us accountable for whether this direction is the right one.
The honest answer to most "will you add X?" questions is: we could. If you ask, "Do you want this in Hytale?"- the answer is more than likely yes.
Right now, we have a list of high-priority, foundational problems that must be solved first. While doing that, we'll continue picking up smaller community improvements where it makes sense - often while we're already performing "surgery" on adjacent systems.
For modders specifically: please work with creators who can help surface concerns clearly and constructively - people like @sly grotto or @native idol. Form a focused group, articulate real problems and real needs, and bring that feedback with a unified, constructive voice. That is how we make this better together.
Maybe it's as efficient as cpp because people write terrible OOP and RAII "modern" cpp code which destroys performance
Or just the fact that JIT/runtime compilation can outperform basically anything that isnβt hyper optimized bare metal C or assembly, and considering we arenβt trying to write a video encoder the performance from Java along with its mature dependency ecosystem (gradle/maven) is a very good choice to build a modding platform on
Eh not really, when u jit u don't get as much time to do more fancy optimizations, so it'd always be less optimized than the equivalent C code using LLVM, also not to speak of the GC slowing it a bit more, there is def a significant performance hit from C vs Java.
U can't argue it's not significant or doesn't exist, u can argue that it's a worthy tradeoff but that's debatable
Also video games are one of the most demanding industries where performance actually matters, which is why things like ECS, game engines, etc are usually written in C/cpp
Nah yeah thatβs what Iβm saying, sure (well written) C and ASM are going to be faster, but it doesnβt really matter because this isnβt a video encoder, itβs a game server that does not need that level of optimization, and something with the simplicity of Java, along with a more mature ecosystem than smth like rust or go are big advantages
Well for the client to run sure, but it sounds like mods are mostly if not completely server side
Which isnβt doing any graphic rendering so π€·ββοΈ
U r misunderstanding, sure we don't need ffmpeg level of optimizations and write assembly, but u can write C, and it's not really that hard and still get something between hand written assembly and java
I do plan on making a modified version of the server with mixin support
Hytale will have a built-in mod library in the client right? What kind of web tools do you think we need? My experience is in web development/minecraft plugins/etc but honestly i'm going to be heavily relying on others for mods and may need to do some of it myself with what we want to do
So that will be fun :3
java is open
Actually servers need to scale by players, servers run all the simulations, and they do so at least 20 times per second (in Minecraft, but other games sometimes do it way more often)
It can be the difference between 100 and 200 players on the server
Also not saying their decision is bad, java was probably the right choice due to familiarity in the Minecraft modding community, but it is definitely not "performant enough"
A Fabric dev told me that it would be a horrible idea but I just kinda want to mixin into random stuff
Do they plan on making a hytale mod loader?
Obviously not
You don't need a mod loader, you will be dowloading mods just like you download mc resource packs
Besides. I only know a singular fabric dev.
Mods are heavily regulated so you won't be downloading rats or anything. They are strictly HYTALE API
And I wanted to mess with mixins specifically
The client cannot be moded π
you will be downloading an asset pack basically. The server holds the functionality*
do we have a date of when the server software/documentation for it is going to be public so we can start working on servers or will that be the same day as launch?
mods are server side
Meant a plugin loader, unless thatβs planned to be supported by hytale themselves
What do you mean by plugin loader?
Something that loads plugins serverside
I meanβ¦ Minecraft is also still mostly single threaded and has been able to run pretty goodβ¦ imagine a well written game engine with concurrency in Javaβ¦ but yeah itβs probably due to familiarity that they chose Java.
Well of course, the whole game is made of multiple plugins π
It'd be hilarious if those are hardcoded with sh**code to rival notch's
Minecraft has horrible performance, I wouldn't use it for comparison
We get the server.jar and documentation on launch day, but there will probably be a hosting guide a bit earlier
sick. thank you
Slikey did post some code before. If you go in the searchbar and search "from:slikey", you can go in the history to find the pictures π
Another thing a lot of people overlook is the sandbox-ability of Java, I couldnβt imagine trying to write a C API that wouldnβt be trivial to accidentally leak memory addresses in, compared to Java that actually has fairly good tools to sandbox (jar in jar, Java agents etc), but I also might just not know enough about Cβs βsandbox ecosystemβ
Btw not true, only parts of it are plugins, most of it is hardcoded
Eh. I don't think it matters much here besides just being a bytecode target
Well idk most, but that was my interpretation of Slikey message
The nicest things about C# and Java is just that the bytecode output is somewhat mappable to the source code and modifiable at runtime.
Thing hytale doesn't leverage at all cuz no mixins
Which is probably why riot was moving to a more performant language like Cpp?
correct, everything is a plugin, some ar jar's and others are just shaded in but you can compile them out.. its far from perfect atm
ultimate goal is that yes the server core is basically just a plugin loader with basic network stack, protocol, ECS and logging / profiling
Yes, they still have hardcoded stuff, but like 1-2 months after the launch? I just feel like it's useless to say "Well actually, some of it is..." when it might not be the case 3 months after the launch
Yes and no. The server is composed of around 70 modules / plugins. There are still many things in the "Core Server" which we are trying to pull out to increase modularity but it takes a bit. A lot of systems are in plugins so that you can simply disable them to save performance if you don't need those subsystems for your minigame / server usecase.
@loud raft
the modding system is the combination of
Plugin Loader
Asset System
ECSthose together allow complete composition for pretty much everything on the server
That answers my question nice
On the server u don't really sandbox u trust it, plugins can just write to files and shi, same with spigot plugins in Minecraft, so no need for sandbox, on the client I'd recommend lua so it's actually fully sandboxed
Thank God it's ECS
That must be really nice to hear considering your earlier worry about notch code lol
You bet!
Based on "many" I assume it's most, and I doubt it'd change much after 3 months, I feel like people are too optimistic for Hytale
Now all that is left is for the community to grasp and understand what an ECS system is like lol
Everyone is eager to move on from Minecraft so Hytale isn't optimistic enough imo
The community when the game doesn't run at zero FPS:
[Parrot eating cracker meme]
Wdym Hytale isn't optimistic enough?
Entity Component System (ECS):
- Entities are just IDs
- Components are data attached to entities
- Systems operate on entities with specific components
This means: - You can add/remove components at runtime
- You can create new systems that react to components
- You can redefine behavior by changing composition, not code
- You can build complex logic by combining simple components
ECS is basically Lego for game logic.
I mean, what I plan to do on Hytale's release is:
- Try the exploration mode
- Once I'm done, try some building (I really want to build trees!)
- Build a parkour map
- Build a dungeon (or multiple)
Is that reasonable π ?
π§ A simple example
Letβs say you want to create a flying wolf.
Asset System
- Define a new wolf variant
- Add a βflyingβ component
- Add a βglowβ component
- Add a βloot tableβ asset
ECS - A βFlyingSystemβ moves entities with FlyingComponent
- A βGlowSystemβ handles glow logic
- A βLootSystemβ handles drops
Plugin Loader - Loads your systems
- Registers your assets
- Hooks into events
You never modify the base wolf code.
You just compose a new creature.
Wasnβt this part of the βnew engineβ developed by riot though? I didnβt think that was developed by the old hypixel team?
Thanks gpt
i use copilot for low-thinking stuff lol
I asked Slikey and he said java version also has ECS, although he said it's way less optimized than flecs
They had a custom ECS in their old engine. They used FLECS in the new one because it's better than what they can reasonably come up with
Ahhh ok
Not sure what ur point is
I feel like people are too optimistic for Hytale
I know it's not what you were going for, still thought it was a little funny comment and somewhat relevant π
There is no actual game loop afaik, their main aim was very good modding at the start so there is what to do until they can get a playable game but their modding is mediocre imo but for some reason people give them the benefit of the doubt and think it's more advanced than it probably is
Someone told me they are aiming to integrate a visual scripting platform which is cool to get newer devs involved
You seem to be the one expecting too much right now...
They said it'd be early access for a few years, it'd not take 3 months to do that
3 months, 1 year, it's all the same
I think it's a waste of time, their direction isn't great imo
How so?
Making a good visual programming language is a ton of effort, effort that should be spent on the actual game
Different people can work on different things. Spending time to make a tool for non-programmers to be able to make mods is not a waste of time.
To make a good one it'd probably take at least a year of full time work by at least 1 guy, that's 3000 hours u could've spent improving other stuff
I disagree, the biggest draw to Minecraft is the fact that it is easily sandboxed, I started programming 7+ years ago because I wanted to make Minecraft plugins, scratch inspired countless developers to learn real coding back in its prime, and history has shown that making coding accessible in a fun way (to mod a game) can inspire generations of developers
modding is crucial. Creators that don't have coding knowledge need to be onboarded ASAP
I think itβs an invaluable asset to the game, and could be what sets it apart from anything else in the industry
Making a visual node editor is not very difficult to do... Sure the nodes behind may take some time to make, but I doubt it would take a year to do.
Slikey said the first visual scripting will come in 4 months after launch
Much faster than 1 year!
He was replying to me in the discussion thread
Assuming flecs is a cpp library. Why not use java 21's foreign function API and just build a JNI wrapper around it?
Yea sure I also got into this from Minecraft, but Minecraft doesn't have a graphical scripting language, and look at it, I don't think having it is any benefit, there are plenty of ways to teach normal programming in a fun way (see boot.dev for example, basically dnd for learning to code)
I don't know! Maybe because they already have an ECS that works. Maybe they would do it but it would take too much time and changing too much
If u want a language like scratch equivalent then it'd maybe take more, designing a programming language takes time (I Kno this because I am working on one)
Everything is third party and it's very hard to understand!
Wow dnd for code, itβs almost like itβs an innovation in the industryβ¦ doing something creative that no one else has done well beforeβ¦ itβs almost like a visual code editor for a video game is innovative and new to the industry lmao
I'd be so happy if they dropped most of the JSON for just a scripting language like Lua
A visual node editor is NOT a programming language
And nuked mcfunctions from orbit
mcfunctions most complex shi on planet earth ong
What is the difference?
Yea exactly, that's what I want as well
I disagree
What limitations are we gonna see right outta the gate? We gonna decompile server software to understand how they've done it?
It's just a structure of a bunch of components that go together to make a result.
In fact, we will be able to make custom node components with Java.
Are there conditions and loops? Then it is
Because JNI code is actually so miserable to deal with, and I think it would generally just be easier to write a new ECS engine than to try and deal with native functions, especially with something as crucial as the entity system π
Well, an AST is a series of connected nodes and that's step 1 to compile anything
It's all just Java under the hood, that's what I mean
I did one in 3 months for a school course solo, if I can do it alone in 1 semester Iβm sure a team of skilled devs can make something good in under a year π
-# ackshually u do tokenization first, but usually ppl put it inside parsing
ππ
Not really, languages usually take at least 5 years, usually 10 to get good and usable
Didn't want to interrupt the ongoing discussion here but just posted something to #discussion that I know some of you in here were interested in.
It isnβt a going to be fully Turing completely language with a dependency ecosystem, concurrency etc π it just needs to be able to interface with the hytale API
I mean... If it's just a wrapper for a glorified builder pattern
Thatβs what Iβm sayinggg lol
Ty, very interesting
It is useless if it's not Turing complete, the other stuff doesn't need to be in the language itself but rather it's standard library (or in Hytale s case, API)
8gb for 10 players π
Also, you don't need a standard library for a visual scripting language like this
Thats recommended
You're just connecting components together, if you need actual business logic go write a new component
So i mean efficency and updates should be better over time right
Yes
Bingo
Well u do, u just call it "built-in nodes", obviously u need less stuff, but a standard library isn't such a time waster, it's usually mostly the language itself
From the sounds of it, its essentially just going to be a script generator that will transpile nodes into simple predictable functions
In which case you can build your node tree and then switch to Java and use its standard library to tweak functionality not available in nodes
At least thatβs how I would imagine it would work
I wanted to build that for #1441742489533943809. The idea was to be able to build logic circuits with a little IC designer table that would produce an entire redstone circuit in block. The catch is that I wanted to compile that to WASM and run it asynchronously from the world.
Scripting language was for client side stuff iirc
So I would take a collection of ANDs, ORs and NOTs, simplify it to hell, then build a wasm module that does the thing and compile it.
The other guys told me it was a bit much for our scope lmfao. They are right.
Still on my list for later tho!
800go ram for 1000 players
Scaling won't be linear
Thatβs if it scales linearly
Sure
You'd probably run into compute issues way before
Imagine if it's quadratic
But its really big compared to other game servers
Might be, if there is some for each playerA for each playerB
Or exponential π
A power of 2 for each 10 players π
Slap an AMD epyc in there and youβll hit thermal limits before either of those issues lol
Anything you can say about what is driving RAM usage up so much? Is it simply chunk loading or is world generation itself using a lot of memory during computations? As your message sortof implies the latter, the way I read it at least
Luckily the Minecraft dev community is known for hyper optimizing anything and everything they can lol
We should hijack the openAI datacenters and put them to an actually good use
Wonder if someone could come up with a fivem like system to handle 1000+ players
They r mostly GPUs tho
(Also can servers do pre-generation to lower runtime RAM usage)
Just answered a similar question in the other channel. Our stress tests were focused on the worst case what players can do, which is chunk generation in multiple independent areas. It might use less without world generation, but that's not validated.
We can use those to run hytale
My bread and butter running minecraft servers on cheap VPS's for my friends lmao
FiveM use a bucket system to split players
Instead of having 1000 players, you have 128 players per buckets and only players in the same bucket can see each others. So you save a lot of performance, because you dont sync players who are not in your bucket
π€
I mean this is what hypixel does with instancing
Il y a t'il une posibilitΓ© de pre generer les chunk ?
No lobby holds more than 100 players concurrently
Is there a possibility of pregenerating chunk?
I have a very important question, is it fun? :3
Itβs different with fivem no? Everyone is technically in the same lobby but you only see 128 players
I actually do not know that. But if it isn't, I assume it will take a very short amount of time for someone to build a plugin for that.
You can setup your own player/bucket limit
We should have a server option to do that
I think π€
The vast majority of my play time is trying to get a server to crash, but I had a lot of fun building my lil house, exploring some caves and being scared by small lil critters in the dark (I'm a baby)
Ok now Iβm confused, how would this help performance? Why place 1000 players on the same instance and use say 100 players per buckets rather than just having 10 instances that hold 100 players?
Is it saving on the overhead each instance has? I guess??
I'm confused by your confusion
You have for example 10 buckets that own 100 players
Only players in your bucket are synced. You dont sync players who are not in your bucket. So in this case, instead of syncing 1000 players, you only sync 100 players (players in the same bucket as you).
If multiple buckets are nears of each others, you sync all near buckets
The idea of how FiveM works just doesnβt sound like it would help performance
Any chance you can do a "stripped-down" benchmark AndrΓ©? Just trying to see what the load is of a bunch of people vibing in something like the Hypixel lobby in size. Try and see how many players the game can handle for just a chill hangout rather than full gameplay, (no chunk loading/worldgen) which could be cool for tons of things.
My first explanation was bad
Thats a pretty cool way of saving performance
Why not?
Oh so players are moved dynamically between buckets based on their location to only render/sync other players in their region?
They can scale Upto 2048 players afiak
Exact
Sounds like MMO region management.
Its almost the same principe as the chunk system. You only load chunk that you see / around you
You dont need to load 10 000 chunk forward :p
Oh I guess thatβs pretty cool, but only really applicable to open world servers and not too useful for smth like mini game servers
Hi all, is there a channel for server-admin recruitment? Like a place to join up with others making early servers?
You can always use the bucket system for mini games
If you have a mini game server, you can hold 1000 players on the same server, but only sync players that is in the same mini game as you
This is an interesting scenario. I will suggest it, but will say that it takes significant effort to pull a very large amount of people together for an internal test, and we have to be pretty deliberate about the scenarios we test. On the Nitrado side we are currently focusing on the scenarios that are most realistic for our typical customer servers.
Make em chunks tiny tiny! π
I assume you don't have client code otherwise you'd just use a swarm of bots?
Are bots testing really the same?
Ok at that point just make a new instance though π all the players you are going to see are already going to be in render view so no need for it
Well, but this has other ramifications like same world state.
Or redis world sync!!
Distributed hytale server
Damn I should really have put the word ramification in my post
Ya now thinking about it. 1k players is a ramification of its own...
You just want a whole bunch of clients spamming the server with random requests and sucking up different parts of world data
Seems pretty simulatable for a server lobby right?
Yeah your right!
Oh? What am I missing?
Nothing as far as I am aware. I just didn't use my brain when asking that question.
Happens a lot...
Thought it was sarcasm lol
A bucket system be a cool endeavour for hytale π€
I have no distributed systems experience tho so itβs out of my skill set lol
My distributed systems experience is bringing down a Minecraft server I was in 3 digits worth of times.
And getting my own database nuked
Losing data one way or another must be the first experience whenever dealing with any kind of server management
Dont worry. We make this up as we go!
There is a German joke about distributed systems that can't be translated (Verteilte Systeme sind eventuell konsistent)
YHuh.
lmfao
I never understood German humor π
||I cant read it, so I have no clue what it says||
i think the joke is that they're basically never consistent
but it is not one i've heard before, so it could be something else π
yea same
it feels similar to when someone tells you to buckle in, because it "should" prevent you from flying through the windshield
One day I will make a distributed game server and they all run lock step deterministic - One day.
German. Joke?
Is that a pun on eventually and the word actually meaning maybe? π
Andre explain your germanness
Specifically the idea of "eventual consistency" yeah
Never
My German is from books and flirting and nothing in-between π
Peak german response
German developer to an english one: Please use the actual password.
English developer: Wait... was I using a fake password before?
Tweet from Slikey:
Yeah, inside of the world configuration, you simply set the pregeneration region and it will ensure that this area is generated before players can join. This should definitely reduce runtime world generation overhead.
i wonder can we customize how the world generates or no tools for that yet?
Yes we can.
We're building Hytale with modding at its core. Most of what you see in the game can be changed, extended, or removed entirely. Blocks, Items, NPCs, World Generation, UIs, systems, and behaviours are all driven by data and code that you can influence.
Data Assets - JSON files
Drive gameplay behaviour and define core content:
Blocks, items, NPCs
World generation
Drop tables, loot, and more.
An unfinished internal editor we used as a temporary solution for world generation, creative tool brushes, NPCs and hopefully more soon.
You can see a picture of the node graph editor in the modding blogpost.
oh ok cool
swedish humor too, because you have to put their jokes together to get them
Hopefully there is proxy support. For networks to be built. Really looking forward to making a high scale server.
a
Even if there isn't one natively. I think we can use other networking tools to allow doing this as well. Maybe with some modifications like networking plugins on the software. Well we will see ultimately. I do hope it's also build into the server jar but we have to see and hope for the best
They natively support transfer packets so for most use cases you don't even need a proxy
Can you show me where this was mentioned Iβd like to research more as I plan on buying a few dedicated servers and running docker networks so I wanna be sure before I start purchasing servers
I wouldn't say for most use cases
If you want to do load distribution and stuff, you still want one
It's just a normal proxy like Nginx, you don't need a special hacky one like we did with Minecraft
Has the server side been tested for performance or are they expecting everyone to horizontally scale. I assume they expect everyone to horizontally scale when it requires dev ops and the community is made of Minecraft Java devs
Can nginx do QUIC?
They are testing for performance. 70 players so far!
Only 70 players?
Well it is a Voxel game written in Java. π
The bad performance for voxels refers to the client side, not server side as much
No it applies plenty to the server.
Sure for some things like physics
I think it's important to note that performance tests are being performed to find bottlenecks and resolve them, and that results may change with every iteration
70 players for a new made server is just bad though.
Hopefully they get a grip on it, so all these mc java devs donβt need to learn dev ops
World Gen, chunk management, simulation, voxel data propagation. It goes on. Its not simple.
70 players in a specific use case. Here, from a Slikey tweet:
I think we will defintiely see 200 player minigames such as Mega Walls or Super Sky Wars. While vanilla SMP is going to require some beefy hardware to push 100 players in a single SMP.
why does it have to be bad?
I would be curious how the packets would work and stuff, and if we can even use a third party add-on to offload stuff when needed. Not that we should have to, but it allows for even more scaling when you can hack the protocol together. Although minecraft is not the best example but the closest one I can think of, I once did an experiment to offload certain serverside computations to other processes, and even gpus in the end. Not that it was a masive success as it was all hacked together and more a proof of concept yet it was interesting. I actually learned that increasing the tps beyond 20 had some interesting consequences on packet rendering. Which I could then use to render ui in a different way and abuse it to capture arbitrary key input. But okay this was all experimental and wouldn't be published in a real project
Because we want good servers that hold many players to minimize developer work
The full tweet:
I wouldn't call it a blog post. It is more of a technical manual with some example data. Just wrapped up a playtest with 70 people on various hardware until we finally crashed the server.
There is a lot of context such as "minigame server" vs "exploration mode SMP". We believe that a 3 vCore + 8 GB server can sustain 6-8 players. A Ryzen 7950x with 12 GB RAM server, we were able to push 70 players but TPS degraded until the server crashed (~45 minutes).
I think we will defintiely see 200 player minigames such as Mega Walls or Super Sky Wars. While vanilla SMP is going to require some beefy hardware to push 100 players in a single SMP.
We are continuing our effort to improve stability and further add performance optimizations to get those big anarchy servers.
Yet it does show the advantages of knowing the protocol eventually even if it means hacking together your own launcher or extending one
The game isn't even out man
we wont be able to modify the client or making new launchers for hytale.
Lol, 70 players flying at max speed generating at the same time is quite imptessive
idk tbh, we have to see released docs.
i hope they give us built-in tools to pregen areas FAST.
Already the case
I dont see why not, they could disable gen all together hand have static defined worlds
Yeah, inside of the world configuration, you simply set the pregeneration region and it will ensure that this area is generated before players can join. This should definitely reduce runtime world generation overhead.
cool thanks for the info
But as for performance, 70 players good start for a very very early access game.
I feel like we should consider the true early access to be in like may/june. Anything before that is like an alpha
(I say may/june because Simon said they would ideally push back the release 3-4 months to get something decent before release)
In a sense, we are getting an early build to make some mods / plugins with before the true "early access" releases then :p
Kind of
It's like "Here, you can play with it while we work on it"
Also we did some comparisons.
We're now at double the average paper enhanced MC server on just your average Hytale server... like damn!
Nooooooo! Hytale server bad π
Not 1k! Need more!!!
We need 1M player civilization in Hytale!
I have a crazy idea to make Hytale's 1st ever Creative mode server where people can build anywhere and edit the world with the tools
THAT would take quite the server lol
Yeh nginx[.]org/en/docs/quic.html
"It just works"
(No Todd Howard gif rip)
fifteen times the detail! See that mountain? You can go there!
Twohundred times the bugs!
The robot can even say your name!
Like offloading chat and other minor stuff into their own services?π€
not chat in my case, it was more like entity rendering
Not yet
Oh like distributed microservice architecture for games?
and I did an experiment with client side shaders which I could manipulate using maps in minecraft in which I could create custom GUI, and more like this but it needed pre rendered stuff which I offloaded to a server with a gpu Woops
In case nobody answered, yes it can
Ah youβre referring to the client, nvm then I was thinking you were talking about the server
nope server
This reads like http3 support, not raw quic
but I did client side stuff using the server side
It does both
it was quite hacky though
I feel like I lost the context of the convo. π something about packets and offloading entity computation?
Interesting. I'll have to read up on that
yeah entity computations, but also some rendering like real rendering, but it's to advanced to explain in a quick chat. I do get the confusion though I should have stated it better
Server doesnt render, and I feel offloading computation as a service would likely be too slow.
If you're interested I can share a bit of my experiments in dms, as I did rendering from the server but mainly to bypass some mminecraft restrictions
Yeah!
as I found a way to send arbitrary data to a minecraft client which it can make sense of
Sounds interesting
I got dis privacy settings so friend!! π
It's possibly only really capable of being done with UDP stream forwarding, but the end result is about the same
Question is if you can match on the SNI field
I just skimmed the source of Nginx and it does look like it supports raw QUIC, but I could be mistaken
I could try to configure it, though I'd need to find a QUIC application to work with first lol
For what it's worth, this is a quote from Twitter
And Slikey uses Nginx as an example
Ah. Well Slikey is a smart dude if I've ever seen one so he's probably right.
If it doesn't support QUIC streams, then it has all the ground-work and would just need someone to tack it on top
Though given how big and famous Nginx is, it'd probably take a long time for it to get merged
Even if someone did do it nicely
I really wanna throw a Golang project somewhere into this ecosystem and a quic load balancer could've been nice π
It would fit in perfectly
The server networks on Hytale are actually going to feel so elegant compared to the awful jank we've done in Minecraft π
It's impossible to convey how awfully janky it is to abuse world-loading and dimensions and use various protocol hacks just to get something so simple in the end
thats actually really nice
I know right, it's going to be brilliant LOL
You just use the transfer packet to send them back to your proxy, and route them according to the 4KB payload
Or you could send them directly, to entirely different networks
It's just a matter of whether all the servers are public boxes, or if they're hidden behind the proxy
I still have a redirecter on my list of things to try. I just need a long evening and a nice cup of cocoa to find out how much of the default server logic I can throw out until the transfer packet stops working
honestly im really hyped to explore the how the server jar works
probably what ill do on d1
I'm probably gonna chuck myself into the world actually
Though I'm immensely curious how creative mode works on the backend lol
fair
I'm seeing a lot more technical talk than what that faq I saw on the hytale blog would allow for, is there some hidden faq I can't find or someone I need to follow on twitter for this info
check out the faq in the pinned messages on #discussion
there is a hytalemodding xyz website where people are tracking a bunch of the stuff mentioned
following on twitter is a great way to get instantly updates and not wait for updates on hytalemodding or something like that
Oh I feel like I should to talk to those folks
Follow Slikey on twitter.
Want an invite to the discord?
btw its hytalemodding dot guide and not xyz
they changed it
xyz still redirects
ah good to know
I can't even keep track of what we say, so glad someone is doing it
Quite a lot probably? I can keep moving the transfer packet even earlier. lol (as long as its safe to do so)
Yeah it already has some of the random plugin examples Slikey posted in the discord the other day
Fun fact, when I was last playing around with that, zero messaged me because I was causing exceptions he hadn't seen yet
@strange tapir Do you guys use something like Jira/Trello to track internal bug tickets? I can't imagine how much needs to be done haha
Me: *what's this crash report I'm seeing... oh its just @vernal niche messsing with stuff π
Yeah we use a combination of things to track work. It depends on different people and the type of problem as to how we track it.
Like sometimes we see a problem and just do it without any need to track it, and other times we have longer term work that goes into a backlog for the future
Its probably better to reach out to the support or creator email
Its on the hytale com about page
Found it, thank you.
There will be a hosting guide some time before early access release
Transfer packets in Hytale are small data payloads (4kb) that allow players to move from one server to another while carrying information. They are not equivalent to something like BungeeCord or Velocity and are primarily intended for smaller servers.
This is pretty opinionated for what hytalemodding.guide seems to be π
It's a stretch to say they are primarily intended for smaller servers
Maybe I'll end up joining their Discord lol
Visual Scripting: Replaces command blocks, fully sandboxed

It was actually confirmed that we'd get developer comments in the server source code?
Yes. It will be shared-source
I figured they'd strip internal comments
We are releasing the server as shared source - you will have the source code with no obfuscation and all our comments in it. You can run very heavily modified versions of the server. If you really want to and you have the technical skills, you can make your own Hytale server software. I know some Rust programmers are already warming up the keyboard. You have to make sure you keep the protocol compatible with the latest version of the client though.
Yeah idk why they claim that transfer packets are intended for smaller servers. Slikey said that they are meant to replace proxies like Bungeecord specifically for Hypixel minigame network-like usecases
Yeah, like it's not replacing Velocity's inter server communication
It's replacing the fact that we need a dedicated server that has to stay in spec with the game and imitate a real server 24/7, just to move people around
We'll still need proxies if we want to do things besides transfer players to public servers, like load balancing and transferring to backend ones lol
Those kinds of proxies are extremely cheap to run though, and are very common
You could grab basically any QUIC proxy off the shelf and use it for your Hytale network
A single-board ARM computer like the Raspberry Pi can easily handle hundreds of web requests a second
We'd expect gaming to be more intensive, but not thousands of times more intensive lol
According to some DigiKey benchmark, the Pi5 hit 100k requests a second lol
Maybe not a safe assumption, because I'm not super well-informed on how request numbers translate into gaming
But I would assume that you're not going to be sending more requests than the client ticks, so for Hytale it would be 50 a second
And that's not even close to saturating it, even with more than 200 players, for a single-board computer lol
Good evening, everyone. I haven't followed the chat much here, but how are the speculations and news about server development?
great so far lmfao
woah, they confirmed linux support has been added. i didnt think that would happen before release. what absolute madlads
It's been discussed a couple times, but I didn't see it posted here
Slikey did end up confirming builtin pre-generation
Yeah, inside of the world configuration, you simply set the pregeneration region and it will ensure that this area is generated before players can join. This should definitely reduce runtime world generation overhead.
Okay, but what I mean is: the architecture and everything, will we actually be able to develop once it goes into pre-registration? Just to analyze and so on. Another question: is there any other place for developers?
We already reached pre-registration, and no we don't have anything to develop with
This is the channel for developers
I mean on other servers.
To my knowledge, this is the only official Hytale discord
Thereβs a hytale modding server, hytalemodding[.]guide should get you there
Is there already any documentation on how to start building servers? I want to start working on it.
Not yet
i just realized their system requirements states this:
-# A graphics card with up-to-date drivers supporting at least OpenGL 4.1 is required. Future versions of Hytale may additionally require support for Vulkan 1.3 and DirectX 12.
I don't have a dedicated graphics card
I actually don't even have a computer right now so I guess it doesn't matter anyways
it's optional
- OS: 64-bit Windows 10 (version 1809), Windows 11
- CPU: Intel Core i5-7500 (or equivalent), AMD Ryzen 3 1200 (or equivalent)
- RAM:
- Singleplayer, with dedicated graphics: 8 GB
- Singleplayer, with integrated graphics: 12 GB
- Multiplayer-only: 8 GB
- GPU:
- Integrated:
- Intel UHD Graphics 620
- AMD Radeon Vega 6
- Dedicated:
- NVIDIA GTX 900 Series
- AMD Radeon 400 Series
- Intel Arc A-Series
- Note: A graphics card with up-to-date drivers supporting at least OpenGL 4.1 is required. Future versions of Hytale may additionally require support for Vulkan 1.3 and DirectX 12.
- Storage: SATA SSD with 20 GB of free space
- Network: 2 Mbit/s connection for multiplayer (UDP/QUIC compatible).
i emboldened the word "integrated" in both places it shows up
your graphics card can be integrated, it just has to have 4.1
Cloud the magic compute machines.
downloaded the last month of slikey and simon's twitter replies so i can search them π€£
i now have a hytale encyclopedia with tweets, blogposts, and outdated client source
Guys can shaders be server side?
You can store shaders server-side, and send them to the client lol
seems like a fun way to crash peoples computers
It won't be available at release because they haven't sandboxed it yet
Why are you laughing bruh.. thats disrespectful as shi bruh..
thank god
Because servers don't render the world, so shaders cannot be server-side
I answered what I thought you meant, that's why "lol"
There isn't an API for client shaders yet, and that's what we'd need
I was just doing the ishowspeed meme
what
Oh i see you unc aight aight
how old are you LOL it's a programming channel
Yeah in that case wouldn't you need to write completely custom server software built for it?
You just need to write Java code that interacts with spacetimedb
then we use whatever interface hypixel gives us to influence the way the server behaves
just how limited that interface is, is still unknown fully
worst case scenario, we hijack the server source fully and not use the interface they provide
It's not a database program, despite the name
You can think of SpacetimeDB as both a relational database and a server combined into one. Instead of deploying a web or game server that sits in between your clients and your database, clients connect directly to the database and execute your logic inside the database itself. No more Docker, Kubernetes, VMs, microservices or extensive ops infrastructure.
SpacetimeDB has enabled us to build our massively multiplayer game, BitCraft, with a small team. Its entire backend, including all game logic, real-time player positions, and all persistent state, is implemented as a SpacetimeDB module.
Ok in that case it would be really hard becaue you would need to get SpacetimeDB to be simulating a Hytale world right?
Not if it can just drop in. As long as it continues the life cycle you can ad hooks wherever for data. So not really anything for a rewrite.
Based on what Shelbie posted, it seems that the hooking into the db defeats the purpose. The database itself is the memory of the server, being updated in realtime, not a copy like an external db usually is.
Instead of deploying a web or game server that sits in between your clients and your database, your clients connect directly to the database and execute your application logic inside the database itself. You can write all of your permission and authorization logic right inside your module just as you would in a normal server.
In summary there are only 4 steps to getting started with SpacetimeDB.
- Install the spacetime CLI tool.
- Start a SpacetimeDB standalone node with spacetime start.
- Write and upload a module in one of our supported module languages.
- Connect to the database with one of our client libraries.
So basically, no it won't be feasible
The Hytale client is not going to be modded, and that seems to be a pretty fundamental aspect of implementing SpicetimeDB
If we assume I'm misunderstanding, and the client-side is not an important part
Then what you need to do is basically port the server entirely to their module system, which would require a rewrite into C# or Rust
And would require a significantly different architecture/design
Much like converting Hytale into a P2P architecture
I think there's technically a way you could do both of those things by making an advanced plugin
So basically all your players would join a "singleplayer" world, which would start a custom integrated server
And that custom server would use the plugin to communicate and sync with the other clients' integrated servers
It's possible that the integrated singleplayer server is going to be more tightly tied to vanilla though
Like, the server itself is open source and we can easily swap out the networking components entirely, even without the plugin system
But for the integrated one, I assume we'd have to stick to the plugin system exclusively
If that's not a problem, and plugins can substitute entire components, then you could technically circumvent the problem and force it to work
If the goal is to get rid of dedicated server costs, I think a more traditional and inefficient forced-P2P plugin sync thing would be easier π
Server only plug-ins seem like they would suck a lot or be really good
I think they're going to be excellent
We've already had incredible feats with server-only Minecraft plugins
And Minecraft is genuinely awful for that, in a thousand different ways
Server only plugins means players don't have to download them manually unless they want to use them for single-player
Cymaera has some very interesting work with server-side only, and no resourcepacks
youtube*.*com/watch?v=uZmEYYs0ZKs&t=119s
In case any of you have never seen it before
And of course there is the Wynncraft team who have done some pretty amazing things with resourcepacks and server-side plugins
The wynncraft team really are amazing
Yeah and we can only imagine how things would've gone differently if they didn't have to wait for Mojang to put literally any work into user content lmao
Also basically the entirity of Hypixel was made using server-side plugins
Wynncraft has been around for so long that it actually didn't even start with server resourcepacks
And even once those were added, you could only texture existing things
And it took Mojang years to even address that part
yes!
Items weren't data driven and you couldn't just arbitrarily apply textures based on NBT either
Minecraft has been literal cancer to work with
They only added it in like 2021 / 2022 I think
The loss of brain cells is prominent enough to get it classified as a carcinogen
But still you can't make native items / blocks. You still disguise blocks
Right now most people use noteblocks for blocks and papers for items
The fact that it wasn't added in 2015 around when they sniped Searge is fckin insane
they added npcs few snapshots ago
lmao the mannequins? genuinely so late is astounding
Yeah when hytale was announced in 2019 I was certain it would wipe minecraft because these guys had given us NOTHING
Still to this day, server software and modding is third party. There's no official help
the most recent snapshot added a /swing command
can you guess what it does? it plays an animation for the target, the swing animation
and that is it, you cannot play any other animation
π
genuinely wtf are they even doing at this point if that's their next move
idk kill the game because they thought they will have no competition
The bar is that god damn low
That's why I think Hytale is going to be fine, even in early access
how is it possible that they have microsoft funding / millions in cash and they cna't hire 5 people to work on creator tools?
wasnt that obvious ages ago when microaoft bought them
Yeah but that was 2014
They just don't even care, they're such jerks
It's more important to profit off child gambling and make it easy to accidentally burn money on the marketplace
Same, people have no idea what it's like working with official API instead of needing 300iq to decompile games and figure it out yourself
id say signs were there all the way back in 2017
It's not even just decompilation
It's forcing things to work in ways they're not supposed to
They were already stalling the mod API for years by 2017 lol
Minecraft Windows 10 edition might've even been 2016
Why? you don't like retexturing symbols? then spawning a WITHER HEALTH BAR and adding that symbol there so you have a custom ui?
i wonder how much resources they are spending maintaining that abysmal c++ port
lmfao they're churning a profit with that sht rendition of the game
right its stuffed full of monetization aint it
yes, you should try it sometime
nah, ill pass
maybe witness their new official server too
it's pay to win by the way
When I grow up I would like to be a minecraft marketplace content creator β€οΈ
they artificially slow your progress with a thousand timers are charge you to skip them, hour by hour
That dragon server is so shi I want to cry
Just wait until you get the ambiguous command: βmoveβ
most of modding is decompiling the game and injecting custom logic into it tho
I laughed irl reading this
almost any unity modding
imagine if they had the balls to add a velocity command, like idk
10 years ago
why do we have to do fckin scoreboard math to multiply motion values
We got forward backward detection systems few snapshots ago too
The game doesn't know jumping from flying or falling, it's all velocity values
It's so easy to cheat too, they have no protection at all
literally anyone could do a better job with minecraft considering the resources available
even a 3 year old baby would be able to hire help and at the very least appeal to other human beings who want to play the game
I'm actually convinced a toddler would do a better job, because they wouldn't be so god damn greedy
facts
Good modding would have an api to work with to provide more content; you should not have to make your own api/hooks to add basic features such as items
Also that 1.9 pvp is so cancerous man
I would literally purchase Minecraft again if their CEO and CFO were replaced with 5-6 year old kids
I play games for dopamine not sit around 3 seconds for the sword to load back up
I agree with this. Mod = modification of the original material
There is a ton of irony with saying this while in an RPG server
If you as a company make MONEY from creators then at least have some dignity and build some tools
Hytale combat is not near close to that...
sorry fella I am using my shield you can't hit me now!
I miss those days man, the og wynncraft was really something
I will be make hypixel skyblock plugin on Hytale π and I will sell it in $5m USD
):
Will we have an official Dockerfile for the Hytale server when it releases? If not I suppose someone in the community will have made this in no time, but that'd be neccesary for using stuff like Pterodactyl
Iβve been looking a bit into the approximate hardware requirements for dedicated Hytale servers. How many maximum players do you think most servers will have (at least at the beginning)? My guess is around 25β30.
there was a talk about this recently lemme find it
i think it really depends on the purpose of the server
Now we need to see what hardware is needed to reach those 56 players. I really don't like that the server is Java, but it is what it is :/
I mean what are the other options? What would you have suggested other than Java π
servers with exploration in mind will have a lower limit of players while servers with maps/premade worlds will probably take less resources therefore a higher playerlimit
almost anything, C#, Golang, C++, Rust
Golang, C++ and Rust would all have made modding significantly more difficult. And I don't really see a reason why having a C# server would be a huge improvement
native aot
AOT for software that has community moddability as a core pillar?
Just because the core is AOT doesn't mean the plugins have to be too. Maybe someone needs a programming course?
Yeah don't get me wrong, I heavily prefer AOT but that's not a great idea for something that is supposed to be incredibly flexible where you can modify every single part
moddability doesn't just mean "having an api". It means we have the ability to easily decompile, edit, and recompile
Yeah good luck using something like Reflections with AOT compilation
Also, targeting the JVM rather than end-user systems is huge for development velocity. A lot of things regarding platform compatibility, memory management, tooling, etc that don't have to be invented from scratch
Multiplayer servers usually use C++ due the performance boost. You can easily adapt sdk to it in any other programming language.
.NET 10 NativeAOT is getting very close to the cpp performance if you consider that C# still has all functionality of memory management.
You would get lower peak performance using 100% AOT in this environment due to servers needing to stay online for long periods. Once you warm JIT you're going to out-perform AOT long-term, especially well-optimized JIT. You can even mitigate that start times JIT produces by caching your startup processes
I wouldn't think so, I'd have a very specific architecture in mind when building and deploying a Java project and ig there are so many options to do it
In the long term we probably see community-made cpp servers when it gets open sourced
I think so too, unless they hit us with the hybird approach and compute is done in C#/C++ and we interact with in thru Java. Tho the memory objects would need to be addressed in a way both langs can read them, which I feel it trivial for engineers at hytale level
Idk Java memory is weird. Unless with JNI, Java low level hybrid approaches I assume would be quite messy
Yeah you'd have to bypass JNI and make java hydrate your memory in a custom structure C++ can read, so we can use java GC and such. We can auto grab the memory java made from C++ to use in Java OR C++ from there. Then there's the issue of concurrency when doing things like this tho
I would be surprised if someone even had the mental motivation to go through that pain xD
"Ok Gemini, convert those files to cpp"
JNI accelerated libraries was a topic some time ago but as far as I remember, no or atleast not many such libraries are used because of equal pain
JNI isn't too difficult to use, I'm just not sure how useful it will actually be. Really depends on what is the bottleneck for some things, if there's one incredibly heavy thing then yes it could be useful. But it won't be too much of an optimisation compared to well written Java code
Imagine when Hytale eventually releases on Steam, first thing you read is "This game was made using AI"
Nah, game files won't be ai made.
Slikey already told me that he uses AI a lot to write debug-, benchmark- and utility tools which are kinda like throw-away software
Yee idk what Steam's rules on the AI classification are really but maybe that counts idk
Then every game is made with AI 
lmfao, i was thinking more in the lines of this: We can use java to make a raw memory object, floats, longs, blah blah. Then we can get this same data using structures and update that same memory objects in both langs, we can tag this memory per domain and time stamp it. That way concurrency is taken care of for newly made memory and we have custom tags to access it. Then we can tag this memory further by hot, warm, cold, etc states and mak a universal registry of the memory both lanuages can use.
I mean isn't that the current controversy? So many games appearing with the AI classification lmao
Yeah, so the core stays cpp but you add a hook to the core elements with small overhead. Should also work with any other programming language, as long the data types are basic
Yes, we must use a concurrent memory registry all langs can access tho for concurrency. Not sure if java lets you make raw concurrent memory
Not using AI, even if this is only for debugging tools, is just waste of time. Ngl. Doesn't mean that the game files are made with it.
You can't fool the players at the end, they feel if something is entirely made by machines 
Yeah but planning is part of making a game. It doesn't sound unsensical to me to say if one uses AI to plan out, design, benchmark, etc. a game then that game is made with AI even if not the code. I wouldn't do it or like it but I could see Steam doing that, for legal reasons alone maybe
Gamedesign needs to be 100% human, a story written by AI just feels so off.
True, atleast that should be the case
I believe AI is a tool, when cars came out people weren't like "this is too good, ima keep walking". I think AI-Assisted is the word you're looking for. Made with AI implies the entire thing was made by AI, like a AI generated image
Chunk pregeneration confirmed π
Yeah ig thats just my wording, I believe Steam says AI-assisted too, I am not sure tho
Bruh, that entire label is just bcs for supporting human made games and blame for those, who did use AI.
We're talking about jobs, nothing else.
But that's the whole point of AI, to make our life easier 
AI is alright if you run a basic model from your pc. But these guys have like 40% of the total electric emissions in a year
I think the costs are way too big for the profits
I do believe a lot of games or work labeled assisted is just pure or 95% AI, that label can be bs.
That's just the tip of it. Remember that AI became common just 2022. Less than 4 years. In this time AI evolved so freaking fast. In the upcoming 10 years, we probably spent 50 - 70% for energy just for AI
Yea but its a technical evolution as basically anything. It's comparable to electric cars, you want to use them to be more environmentally responsible eventually, even if not now. And using them now already is important to the progression. Same with AI I guess. Technology like photonic processors and various other efforts to reduce power output are on their way
I thought this until I watched a video on farming and how much corn we piss away, how much water it takes and what we use it for. I just think its a new guzzling medium everyone is buzzing about because of the media attetion of AI
Infrastructure is still the main problem . We got the solutions, but we humans are to slow for the modern world.
We reached a milestone where technology evolution bypassed the capabilities of humans
It doesn't even generate profit
the US gov holds em together
I don't even think its this, I think its profit > stability. There's plenty of people who LOVE to tell how how bad something is for the environment, but they dont control the laws sadly
Telling doesn't mean you get the work done 
If only lmao
That's another issue, the big companies are just passing around money, gov just building data centers near neighborhoods with 0 regulation. The upper half US eco would be in shambles without AI
green laundering
Government backed green laundering*
ye theyre all in on this, these past few decades so many governments are pushing deep into corruption
remember times when governments were working for the betterment of the people? i wasnt alive back then yet
Just mind to keep politics out of here btw lmao
nah i started it a bit too my bad lol
We were fine without AI bro
I use it everyday as a tool, but I feel you tho. For me, it helped a lot with my return to development and impostor syndrome from the systems I have made. Im a better developer because of it
There's no denying that AI is useful but you have to count the fact that it's bad for everyone
For simplification reasons I always tend to say that AI is not a tool for beginners but for people who know what they're doing lmao
This may be harsh but I think people need to get WAY smarter now and do things AI can't do. In development that's your system level thinkers, team leads, etc. This isn't the first time automation has caused something like this but it is the first time its been so fast, and so vast. It does need to slow down, but the money is too good and we're small pawns man
EXACATLY THIS LOL, I never figured out why people say AI is bad until I really read the responses because I just skip the bad stuff lol
Me when I am a silly little rage baiter
I guess its nice buying gpus 500% more
I accentuated it's as bad as people say it is AT TIMES, but there are reasons for it that I'm aware of. Prompts, context, your chat flow, predefined agent prompts, tools, etc, they all matter in the outcome of your prompt. This is why no one prompt will give the same response even on the same model
Well it's okay. It will be banned eventually
Even if it does this won't stop people from running local models. And a 100% ban is never going to happen, regulation maybe, ban? No. America itself has actual money invested into these companies
Generative AI may get regulated here and there but that won't affect the market reasonably I'd say
Local models is fine, but creating miles of infrastucture and ruining nature is not cool
Inference isn't what's expensive. The training is
Yes, this depeneds how you to do this. The new datacenters being built in the US are no-where-near up to par with a real data center environment standards. There are data centers in other countries with net 0 emissions for entire operating years, water is not poisoned, temperature regulated before returning, and 100% renewable energy is used. When done right, a datacenter should only cost in environment tax in what it cost to get built
It's normal that we face disbelievers on the way down the technology road.
When the first cars were driving on the road, people with carriages said that they don't have power to pull loads of goods.
When electricity was commonly used in public, the inventors of the pneumatic clocks in Paris said that the system will fail to many times.
We are at the beginning of AI, not at the end or even near middle of it
Yes, this is why I can use AI to HELP me build complex systems, not be the conductor. It's too good not to use for development and you're going to be behind if you don't use it.
I am not arguing about the AI philosophy. But we have been victims of the industrial revolution's polution. We know better than this
Its just not our call, so you may as well try to profit or learn from it then deny the changes that WILL be made.
Thankfully I am not American so I don't have to worry about random billionaires deciding to ruin my continent
Where do you live?
Every technology comes with it downsides.
The problem is, that if 99% of all countries will follow the no-pollution way, there is still one country who will do it and this one will out-smart everyone else in technology development
That's def not America π
Issue is there no money in emssion control, just costs, needs to come from the top down with HEFTY fines
So the plan is, to make faster development which leads more to energy efficiency until we meet the break-even point where we don't need more energy for ourself.
This point is WAY far in the future. Maybe in 1000 years we got reliable fusion energy which MAY fulfill our demands
Our globally energy production already could fulfill our current needs, which will lead to more energy consumption.
Currently we invest trillions into infrastructure just for consuming more energy π
DnSpy enjoyer
I donβt see a way where you are able to do transpile patches with NativeAOT
ilSpy supremacy
π
ilSpy mentioned!!!!!
Slikey shared a screenshot of the config for pregen on Twitter and in it, you can clearly see that X and Y is used for horizontal plane and therefore Z has to be vertical... welp, that's going to confuse my brain for a little while.
Source: xcancel(.)com/slikey/status/2001032986695811152#m
wasnt that confirmed on some map screenshot a while ago?
I must have missed that screenshot cause this was the first time I learnt that information
Minecraft is kind of the outlier when it comes to using y for its vertical axis
LETS GOOOO!!!
Hi
