Epic Games' CEO Tim Sweeney: Linux has a 'terrifically hard audience to serve'
theverge.com
theverge.com
>We don’t have confidence that we’d be able to combat cheating at scale under a wide array of kernel configurations including custom ones.
https://twitter.com/TimSweeneyEpic/status/149056592564871578...
People hate cheaters though so that's a really hard problem to solve
“ mess with the os of Linux users like we spy and fuck up users that are using Windows “
Maybe those worlds do overlap but anti cheat measures do seem like something you have to do to keep the game viable / people playing.
It’s a big reason why a lot of people continue to play Age of Empires II via Wine today.
Yes that is why so many people prefer to play on console. Especially when we caught more and more live-streamers cheating.
<<Epic has now stepped in to defend itself from those accusations, while also admitting to an "outdated implementation" that can make unauthorized access to local Steam information.>>
> “Linux is a small market already and if you subdivide it by blessed kernel versions then it’s even smaller.” It may not be worth Epic’s while to put in the work on security for what will be a comparatively tiny audience, at least at first.
For many products, there's not currently an economic incentive to support Linux because the development cost, support cost, and opportunity cost outweigh the potential revenue. I'm sympathetic to this because it's something I face with my own product. In many cases, the Linux market currently isn't large enough to justify the cost, but I hope that changes in the future.
In any case, I don't imagine this will be a long-term issue. The Venn diagram of Linux users and Fortnite players are two distinct circles, distanced six feet apart.
The Steam Deck and others like it might close this gap. I know if I was looking for a gaming PC for my pre-teen/teen, I'd be looking really hard at it.
It can also run Windows though so that might be a good (if relatively pricey) workaround for that subset of users.
* Be pure ELF64 (exe and so), which depends only on libdl (static loading) and its 3 symbols (dlopen/dlsym/dlclose). Pure ELF64 means it must not use "main()" but the SYSV ELF ABI entry point (which is actually almost main()...) and use fine grained linking. * use pthread TLS, and not ELF TLS symbols (do not know if compilers provides "switches" to do that). * window system: wayland(static code)->x11(xcb libs) fallback. * 3D, gfx acceleration: vulkan (GL... mmmh... I would drop it) if the binaries are a game or a GUI which _really_ needs some sort of hardware acceleration (CPU with media instructions are seriously fast for 2D). * input: dynamically load libxkbcommon(wayland) libxkbcommon-x11(x11), that for the client side xkb state machine which is fed with keycode inputs. Joystick support would directly use kernel events and the evdev file protocol. * sound: dynamically load libasound (pulse and pipewire are hidden behind the alsa api, if ever present since it could be a simple dmix/dsnoop).
Unfortunately, c++ being c++, gcc/clanc/etc c++ runtime must be forked to be libdl-ized (since it is not the case), and all 3rd party libs.
This is significant work, but brutal and supposely ez.
I'm quite skeptical about their argument though, since on Windows you can also disable secure boot and then run a custom kernel or a custom hypervisor, so it's only valid if their anti-cheat rejects all Windows hypervisors and also checks the integrity of the Windows kernel (in this case it's still of course not secure, but the anti-cheat itself would need to be patched to defeat it).
What if I want to play with my friends who aren’t on Linux? Oops we removed cross platform capabilities.
Also making all the players who can’t run anti cheat play together means almost any individual non-cheating player will be overrun by the cheaters and essentially pushed out of the game.
You can play with them just fine, but if you use match making you would only be matched with other groups where there is at least one person running Linux.
> Also making all the players who can’t run anti cheat play together means almost any individual non-cheating player will be overrun by the cheaters and essentially pushed out of the game.
Just because people can cheat doesn't mean they will in significant enough numbers.
Also the so-called "cheating" is just a perfectly normal way to use a general-purpose computer to run code of your choosing and to automate tasks, so you can do it as well if you want, and it's perfectly normal that if you refuse to automate your performance may suffer since you are using an inferior strategy.
Games have been abandoned even though companies actively catch cheaters because no one wants to play against someone with an unfair advantage. If Epic releases a game where they literally cannot catch cheaters, that game will be abandoned in a week.
Your defense of cheating is laughable. People are trying to play a game with a closed set of rules. Saying that your hacks are legal because they're turing-complete is completely nonsensical. You are not playing the same game that people signed up for. You're hijacking someone else's enjoyment for your own power fantasies.
If it was virtually 100% of anti cheat would be server side which is much closer to 100% effective.
Please nobody put this guy in charge of anti-cheat for any game I play
Angry Tweeter: “Newsflash: CEO does not trust his own product”
Tim Sweeney: With regard to anti-cheat on the Linux platform supporting custom kernels and the threat model to a game of Fortnite's size, YES THAT'S EXACTLY RIGHT!
We have essentially two cases to consider:
1. The cheater is using a rage cheat to get ahead of everything, in disregard of wether he will keep the account. This is pretty easy to detect programmatically, even more so for a human watching the demo. This case is rather rare, because accounts tend to be expensive/annoying to get into a state where competitve queuing is possible.
2. The cheater is using a tool to covertly rise the competitive ranks without showing his advantage directly. This is also not a problem because for 99,9% of players there is no difference if they're playing against a cheater who gives himself a slight advantage or simply against a better player. This player will rise through the ranks until he finds the spot where his tool assisted performance places him or he will have to increase the assistance to a point where he's getting into category 1 territory and is banned. If hidden cheating is getting important because money is in the performance when reaching the top 0.0001% of players, we're getting into lan territory where the hardware setups are either provided or checked.
For fair players type one is annoying, but a rare occurance due to the deterrences measures taken for new accounts. Type two is moot. I conclude: Cheating is not an issue for normal people.
It used to be a meme that any decent CS player got banned once a day for hacking by clueless admins, and that was with review (CS had great replay tools for server admins) by players who cared enough to host their own servers. That was also in the days where if you got banned, you moved to another server, no biggie.
In the F2P shooter world, there are often several hundred dollars of spend on cheaters accounts, because the people who crave instant gratification of cheating are generally also the whales. That creates a perverse incentive on both sides - Ban then and you've taken their money for "nothing", don't ban them and get more money from them.
1. is absolutely not rare at all, particularly in F2P games and even in paid games. I've seen periods where CSGO and PUBG games were filled with blatant cheaters. They were usually detected and banned within a few games, but the community still had a problem because there were enough of them that there would be a few cheaters in every match ruining the whole thing. The metric that matters for a player is cheaters/game not games/cheater. Given enough cheaters, they can dominate a large fraction of games even if they are banned quickly.
2. is also frustrating to play against. There are cheats that show players where everyone is on the map in COD Warzone or PUBG. It's subtle to detect, but in practice is a huge huge competitive advantage and gives you 0 chance of surviving an encounter with them. And just like the previous case, they may rise through the ranks eventually but ruin 100s of 100s of games in that time. With a large enough fraction of these cheaters, you will encounter a few every game you play.
Not to mention other perverse incentives - the "smurfing" impulse many cheaters have, where they will abandon the account as soon as it gets to a high enough rank go be even remotely competitive. They just create a new account at level 1 and start all over again. So the steady state you imagine never happens, and that cheater goes right back into the pool of low rank games.
Banning an account wrongly is also very costly. A good player wrongly banned will hate your game forever. But a cheater often will have no qualms spending money to buy a new account even if banned. So the detection systems have to have both very low false positives and high recall on true positivies and very quick detection of true positives.
Another thing HN frequently overlooks is "but I could bypass this by just writing a fake mouse driver/run on Android blah blah". Cheaters are generally people looking for a quick hit. They are not the type of people who will spend days or have the knowledge to setup whatever complicated _technically_ possible but difficult workaround. Making it difficult to install, buy or use cheats easily is an effective deterrent.
It's really a very difficult problem.
1. Is and always was rare. During all my time on official matchmaking servers I encountered maybe one or two raging spinbots per year tops. I don't know if this is region dependent, I always play on EUW.
2. Again, how do you know? Do you watch the whole demo after the game to judge wether someone was cheating all along or do you just go "he just couldn't have known I was here, he must be cheating"? If they raise through the ranks and "ruin" games along the way, how is it not blantant enough to cause an ban?
Besides, I don't know much about the fow in those large-map royale type games, but why does the server even sent every other players position to the client? Hiding as much info as possible from the client from the start is like anticheat 101 and in most cases much, much easier than trying to get the players computer to disobey him. In Counter Strikes case, player positions are only sent to the client if they have to be, based on how the maps parts are attached to each other iirc.
> Not to mention other perverse incentives - the "smurfing" impulse many cheaters have, where they will abandon the account as soon as it gets to a high enough rank go be even remotely competitive. They just create a new account
There is two parts to this. The first is the smurfing problem, which I personally find the most annoying thing in competitive matchmaking based videogames to date. In regular sports, when someone joins who is much better, you could always just say "fuck this, I'm leaving", but in competitive games only you are punished for leaving and the other guy gets to abuse you for the duration of the match. I have yet to see effective counter measures against this and I doubt there will be any soon. It is a legitimate player after all, there is no real reason to ban him since bad intention can rarely be proven. Many people abandon their accounts for months, get back worse and climb the ranks again afte swiftly resharpening their hibernated skills. Note however that smurfing has nothing to do with cheating -- it's generally just players who misuse the ranking system for their own amusement while destroying the fun for others.
The second part falls under the same catergory as my original first point. It doesn't really matter if the cheater is banned or just gets bored and makes a new account. This is directly correlating with the friction of creating new accounts and it's not an anticheat problem. Back in the day when CS:GO still cost money, there where waves of new smurfs and generally trolling or annoying (and yes, people I would suspect of cheating as well) players after every steam sale when the game would be sold a neglectible amount of money. Now it's a lot better since new accounts are distrusted by default and require a lot of effort to get into a trusted state where normal matchmaking is even available, dramatically reducing the number of throwaways due to the sheer amount of effort of creating a new competitive account.
So this problem is nothing that couldn't be reduced to neglectable amounts if the makers of the game cared enough to properly filter new accounts.
> Banning an account wrongly is also very costly.
Yes, I'm not sure how this is related to the discussion. If you ban my account because I looked up what cheats are available for a game, sent my entire browser history to the mothership to throw it against some domain blacklist and ban my account based on that I'd be pissed as well. False positives are nothing an invasive anticheat system wouldn't have.
> Cheaters are generally people looking for a quick hit. They are not the type of people who will spend days or have the knowledge to setup whatever complicated _technically_ possible but difficult workaround. Making it difficult to install, buy or use cheats easily is an effective deterrent.
I have never even once heard of a cheater who programmed his own cheat for a game and inteded to actually use it for competitive matches himself. There are two types of people, and both are highly sophisticated:
The first type creates cheat for shits and giggles. He does not intend to actually use them beyond proving that they indeed fall through the filters.
The other type is a professional cheat publisher. He has the expertise, often publishes for multible games accross genres and offers personalized cheating systems on a subscription basis. The people willing to use the cheat will follow every given instruction, even stuff like rebooting windows into recovery to install unsigned drivers and things like that. They have no idea what harm this can do to their systems and they don't care because they are not techincal and don't need to be. It's a pure service economy, and not even an illegal one.
Source: I had a friendish guy for some years who eventually had his account banned because he stopped installing updated versions of a cheat and eventually got caught when VAC got onto the older one. Everyone who played with him had no idea he was a cheater until that moment, he just hit some crazy shots from time to time, sometimes while he was high on weed. Enemies didn't know nor complain, neither did we. It was indistinguishable to a high-skill player.
> It's really a very difficult problem.
Yes, but I'm convinced Valve solves this with great brilliance.
Next time you tell someone he has no idea what he's talking about, please consider you might be wrong.
It does ruin the game for me, and Fortnite has the same issues as it's another free to play battle royale. Anyways, I'm just saying cheating is a real frustration for some, it's a different situation in battle royales than in CS, and some of us would rather run a kernel module to help catch them than not. I don't think it's a non issue or can be solved the same way as valve does for all games. I think Epic is right to limit Linux when it could result in an influx of more cheaters, has a low market share, and high fragmentation that makes it expensive to develop for.
Even if they disagree with Epic's reasoning, Epic is free to hold different opinions and do what they like. Overall, I feel it's ridiculous for users of any niche to feel _entitled_ to having a company serve them and get mad when they don't. That is like elevating the ability to play a specific game to the same standards as a critical service like healthcare or a constitutional right.
Sure, but both of these can be checked server side. The server can just tell the client "gunshot with $weapon from x/y/z" without compromising any more information. Just the same with the sniper - player has $weapon and looks with vector x1/y1/z1 from point x2/y2/z2? Now we can send the data of $player2 and $player3. Depending on how involved this should be one can even check for rough line of sight depending on terrain etc. I know that this is very sophisticated server side logic, but client side anticheat requires sophistication of similar complexity. In my eyes, when a publisher tells us "we have to have client side anticheat" it's mostly because they're taking the lazy, sometimes borderline-placebo way (good for you if it works for your games, I'm not saying it's not)
> And it is very clear they are cheating if you watch a replay, because they will e.g run across the map with no fear when no one is around, slow down near enemies, walk into an empty compound and grenade the precise room someone is in who they never spotted.
Sounds like an easy job for some poor outsourced slaves in $developing_country. I agree that checking for radar/walls is nearly impossible to achieve with purely electronic measures. In CS they introduced the overwatch system for this purpose, where reports result in the demo being anonymized and presented to other players who can review fishy demos and determine wether someone is cheating or not. I personally would've liked them to not task the community with this.
> More subtle cases are possible too, where it is possible to be fairly sure someone is cheating only after watching many replays. If it's that difficult for a human, I can accept it may be near impossible for a computer.
So if you can't tell if someone is cheating if you're watching the demo, how is this not falling under original point 2? Apart from the ethical argument of a level playing field and a fair match, how is it different for you if the other guy is just better than you in contrast with him maybe cheating when watching his last ten games?
> I think Epic is right to limit Linux when it could result in an influx of more cheaters, has a low market share, and high fragmentation that makes it expensive to develop for.
> Even if they disagree with Epic's reasoning, Epic is free to hold different opinions and do what they like. Overall, I feel it's ridiculous for users of any niche to feel _entitled_ to having a company serve them and get mad when they don't. That is like elevating the ability to play a specific game to the same standards as a critical service like healthcare or a constitutional right.
I did not say they're wrong either, I gave no comment on restricting platforms that allow for general purpose computing. They can of course do that if they want. My whole commentary is on the ways a provider can provide anti cheating measures and how I think that an aggressive client control is not the right way.
In my eyes, when you have issues with cheaters in your games despite the extremely aggressive measures taken by $provider, you should maybe start blaming that provider for cheaping out on humans to help get their problem under control.
I also don't quite understand why you would keep playing games that are "ruined" for you. I usually play games because it brings me joy, not because I want to worry and get worked up.
> Actually, spoken as someone who has played his fair (1000h+) share of Counter Strike Global Offensive since it came out
Something doesn't add up about your story. You really played 1kh+ and have never seen how much people hate cheaters?
Anyway, there's a very simple psychological explanation for why they do: Video games are only as good as the rules they abide by. If you go into a game and are surprised by rules you didn't know existed, then that's no fun. Likewise, if you have to guess if someone is a hacker to determine if you're able to defeat them, that's a surprise, and no fun. It's frustrating enough when you pull off a daring maneuver and suffer a framerate stutter or network congestion and the opponent gets away or destroys you. Now imagine when the other person (surprise) is playing by a different set of rules where they can see you around corners and never miss.
> if you have to guess if someone is a hacker to determine if you're able to defeat them, that's a surprise, and no fun
Seems like you didn't understand my argument. If someone is indeed cheating and you don't notice anything fishy, how do you know that he isn't just a better player? If you can clearly see that he's cheating, how is that player not banned after a short amount of time?
Pleasantly surprised this is not the case (as it appears now)