00:00 < bridge> Basically a payload that is encrypted code 00:01 < bridge> so how would the server know you have it 00:01 < bridge> It would have a secret key inside it which it would use to communicate with the server 00:02 < bridge> hack clients would just spoof it 00:02 < bridge> (Yes one can extract the key, but there are ways to make it veeery veery annoying) 00:02 < bridge> Not unless they can extract the key they can't 00:02 < bridge> 1. extracting the key from the client 00:03 < bridge> 2. from packets, which would be even easier since its unencrypted ud[p 00:03 < bridge> 2. from packets, which would be even easier since its unencrypted udp 00:03 < bridge> The packets would be fully encrypted between the blob and the server 00:04 < bridge> Basically the client would call into the blob, the blob would independently do stuff and send back an encrypted message 00:05 < bridge> Use TPM like Intel SGX or the newer version 00:05 < bridge> The issue of how secure it is is not really a huge concern of mine. We can make it very annoying, we can generate the blobs daily on the fly with rotating keys. 00:05 < bridge> 00:05 < bridge> The concern is that it would pretty much give us remote code execution on your system 00:05 < bridge> but were waiting for quic to have this ecnrypted communitcation 00:05 < bridge> so accounts would come first anyway 00:08 < bridge> I did think about this (the secure boot research I was doing inspired me). But I'm unsure if we can properly remotely attest sgx. What it reports back to us can be spoofed if I understand it correctly 00:08 < bridge> No, you cant spoof it if you run it in an enclave 00:09 < bridge> Intel SGX had several flaws already with hardware hacking and so on, but if you run it in the enclave you cant access anything besides your loaded code 00:10 < bridge> I should research that a bit more if you say it could work 00:10 < bridge> You need hardware support for it tho 00:10 < bridge> That might be a blocker 00:11 < bridge> Old systems often dont have TPMs 00:11 < bridge> I'd assume most modern hardware supports it by now and amd probably has an equivalent 00:11 < bridge> amd has SEGX 00:11 < bridge> amd has SEV 00:12 < bridge> There is a cool tool for it: https://openenclave.io/sdk/ 00:12 < bridge> So it allows you to run a piece of code in a measured way, right? How do you attest what was run remotely later? 00:13 < bridge> It allows you to load an enclave in the hardware module which is pre-signed & does some work for you. It has limited access to some crypto stuff like generating a private key and so on. You can also use DCAP (Data Center Attestation Primitives) - otherwise you need to use the Intel API = money. 00:14 < bridge> For half a year I am currently researching on how to run OpenEnclave with Zero Knowledge Proof to let other people host KoG servers, but I can always be sure that they are not tampered or somehow changed 00:14 < bridge> Who signs the enclave? 00:15 < bridge> You, with your private key 00:15 < bridge> So when the enclave is loaded, the cpu verifies the signature? 00:15 < bridge> The TEE does 00:15 < bridge> It doesnt verify your signature tho 00:15 < bridge> but it signs the attest 00:16 < bridge> And you have a signer in the attest, which correlates to your private key 00:16 < bridge> What key does it sign the attest with? 00:16 < bridge> Intel on-board key 00:16 < bridge> Its a signed key from intel or amd 00:17 < bridge> Its on the hardware 00:17 < bridge> you cant change it 00:17 < bridge> Even if you do, you can verify the key if its from intel or amd 😄 00:17 < bridge> Okay, so that's the root of trust, that's good. So could we run the entire client in a TEE? Is there a large performance penalty? 00:18 < bridge> No, you are memory limited 00:18 < bridge> You treat the application as untrusted 00:18 < bridge> and load the enclave into the trusted TEE 00:19 < bridge> Hm, so what do you do in the TEE? Can you measure external stuff? Can you get it to sign things with the intel key for you? 00:19 < bridge> Yes, the enclave can be opened within the ddnet client 00:19 < bridge> and it can communicate through a dedicated channel 00:20 < bridge> The client can just lie to it, no? 00:20 < bridge> Depends on the interface 00:20 < bridge> I'd need it to be independently measuring the client 00:21 < bridge> Send a challenge to the untrusted ddnet client app 00:21 < bridge> And report the measurement results in a signed and preferably encrypted way 00:22 < bridge> Yeah, so the enclave creates a private key & then sends a verification request to a public server through the curl implementation in ddnet 00:22 < bridge> You cant fake it 00:22 < bridge> How do I know the enclave created the private key? 00:23 < bridge> With the attest 00:23 < bridge> How does the enclave know that the untrusted part hasn't been modified? 00:25 < bridge> It doesnt by default 00:26 < bridge> anyone can load your signed enclave to create an instance 00:26 < bridge> The issue is what does the binary blob do? There's a lot of ways to cheat even without a self compiled client 00:26 < bridge> I'll just have to read more about this. A simple challange would only tell me that the enclave is untampered with and is running. But it doesn't tell me anything about the untrusted part, I'm interested in verifying that people are running the client they say they are running 00:27 < bridge> The binary blob measures parts of the running executable and reports back 00:27 < bridge> i feel like replay bot / slowdown->replay cheats are theoretically impossible to detect unless you can actually differentiate the inputs 00:28 < bridge> One could perhaps fool it by creating a bogus stack frame to misdirect the blob into measuring the wrong thing 00:28 < bridge> can't someone simulate the inputs in a legit client by using a "fake" input device? i'm not sure if that's a thing in other games but i wouldn't be surprised 00:30 < bridge> yea, simplest way is you would just record a macro 00:30 < bridge> doesnt need to be in game, you just replay the inputs 00:32 < bridge> The hope is always that it's just too expensive and too annoying to do for a game as unimportant as ddnet 00:32 < bridge> There is no perfect solution. Truly trusted execution on an untrusted platform is an impossible problem 00:33 < bridge> Anyway I would imagine the larger issue is political, you'd have to ban any client that you don't control and doesn't have your blob 00:33 < bridge> I would like it to have atleast proof that he runs ddnet 😄 00:33 < bridge> I dont mind supporting other clients too 00:34 < bridge> I was thinking of making it optional. If you don't run it and you are acting like a bot, it's now more obvious 00:34 < bridge> Well you know how the current proof for that is working 00:34 < bridge> Yeah, but its not super reliable anymore as more and more people are finding it 00:34 < bridge> I was surprised the current one lasted as long as it did. It is completely open source 00:34 < bridge> When we implemented it I told heinrich it would take people at most a month to beat. It survived 3 years 00:35 < bridge> Voldemort makes big money 00:35 < bridge> Big motivation 00:35 < bridge> and how long did it took ddnet to ban "bro, check out this client" phrase? 00:35 < bridge> 💀 00:35 < bridge> run a free alternative w better features and shadowban the users over time so it's inconspicuous 00:36 < bridge> Still not banned lol because its fruitless 00:36 < bridge> I should start detecting that again before messing around with other things just so he makes less money 00:36 < bridge> Only the free version does it 00:36 < bridge> ig it still reduces advertising tho 00:36 < bridge> Anyway, I was just curious what everyone thinks about the possibility of a bit of very encrypted code 00:37 < bridge> I dont like it, neither as a hoster nor a player 00:37 < bridge> Idk accounts seem more productive 00:37 < bridge> Evsn if it's optional? 00:37 < bridge> Yes. I dont want to image what happens if any server hoster can run anything 00:37 < bridge> Even sandboxed 00:37 < bridge> Accounts are idk, they got out of my hand, I'm not very happy with how they ended up and I don't really feel like working on them 00:38 < bridge> Just a small security breach somewhere 00:38 < bridge> Noo, that I would never allow. No way. Only ddnet makes blobs 00:38 < bridge> I still don't think the number of current cheaters are bigger than what we can handle by hand, its only because we lack any form of "collateral" that can be taken from a player when they'd banned that mods can't keep up 00:38 < bridge> I still don't think the number of current cheaters are bigger than what we can handle by hand, its only because we lack any form of "collateral" that can be taken from a player when they're banned that mods can't keep up 00:39 < bridge> Then I would be in favor of HWID 😄 00:39 < bridge> Giving all hosters remote code execution is insanity 00:39 < bridge> Instead of ddnet blobs 00:39 < bridge> ddnet rs does this in a way 00:39 < bridge> but sandboxed, no? 00:39 < bridge> It'd need to be calculated in a very secure blob 00:39 < bridge> Yes obv 00:40 < bridge> In a wasm sandbox, what I'm talking about is blobs that have full access to the entire address space of the client 00:40 < bridge> hmm 00:40 < bridge> It would basically be another executable just extremely obfuscated 00:40 < bridge> That doesnt sound good 00:41 < bridge> Will AV like this? I feel like it should be massive red flags if theyre paying attention 00:41 < bridge> But maybe theyre not idk 00:41 < bridge> Yeah, it makes me uneasy too, that's why I wanted to ask around see how others feel. The only saving grace is that people already trust ddnet to not ship malicious binaries. If we wanted to do something evil we could just ship it in the binary 00:42 < bridge> True that, but now we dont want to start with shipping evil things, dont we? 00:42 < bridge> Well if we did they don't need to be in an encrypted blob is my point. We technically already have remote code execution with the autoupdater and steam updates 00:45 < bridge> Anyway, SGX sounds fairly interesting. I might look into it see if I can use a TEE to measure stuff externally 00:45 < bridge> Lets ask Riot Games with their Vanguard anti-cheat 00:46 < bridge> They use some very spicy secret sauce. Their kernel driver unmaps critical memory when it's not needed 😄 00:46 < bridge> I still prefer account based collateral ideas, im gonna keep bringing it up 😄 with HWID even is ok. 00:47 < bridge> The process simply does not even have the critical sections mapped, when they need access the kernel driver provides access only if it still trusts the binary 00:48 < bridge> HWID is just a no go. How are we going to stop people from spoofing it? Unless there is hardware support for signed hwids 00:48 < bridge> Idk your encrypted binary blob 00:49 < bridge> Yeah, you need something like that 00:49 < bridge> Yeah, you need something like that. That’s why I was floating the idea 00:49 < bridge> You just need a component that you can trust that has a key in it that is hard to extract 00:49 < bridge> Yeah but I want the accounts first, then make it harder to bypass ban later 00:50 < bridge> I don’t want to talk about accounts now because I’m still very very unsure about how they ended up 00:50 < bridge> accounts ❌ removing botted record ❌ encrypted binary blob remote code execution ✅ 00:50 < bridge> And it’s fairly late, if we get into it it’ll take hours 😄 00:51 < bridge> Well how accounts actually function isn't in my control I can only hope they'd be maximally useful 00:51 < bridge> But if we're allowing unregistered players to set ranks or assist others then its kinda pointless anyway 00:52 < bridge> top10 should be register only 00:52 < bridge> not even gonna lie 00:52 < bridge> Top 100 00:52 < bridge> yeah, and any hard maps idk 00:52 < bridge> Yeah, we will have something like this in the next upcoming version 00:52 < bridge> Imo the ladder should be split if possible 00:53 < bridge> right so regular ranks and TAS leaderboard 00:54 < bridge> i liked the ddnet season 2 idea w the split leaderboard pts and (maybe map pool?). idk how else you'd go abt a split ladder 00:54 < bridge> There are 2-3 solid proposals on detecting TAS runs but I think the people who proposed them are not working on them 00:56 < bridge> what are they 00:56 < bridge> Detecting duplicate ranks or detecting TAS? 00:57 < bridge> Idk if I’m allowed to share them. Detecting duplicate ranks 00:57 < bridge> detecting tater 00:57 < bridge> :brownbear: 00:57 < bridge> I gave lots of people teehistorian files to make it happen, no one got back to me with something that works :pepeW: 00:58 < bridge> It’s sleep time, gn everyone 00:58 < bridge> gn 00:58 < bridge> https://tenor.com/view/pipotam-baby-hippo-hippo-wash-shower-gif-3406421565683572068 00:58 < bridge> do most tasers duplicate their runs on official servers? 00:58 < bridge> seems kinda dumb 00:58 < bridge> i guess people share their replay files 01:02 < bridge> i don't see how detecting duplicate runs is too hard of a problem 01:02 < bridge> 01:02 < bridge> detecting single tas runs is interesting though 01:46 < bridge> you should get like 1-4 rank points if you are top 50%-1% 01:51 < bridge> we had that but it was hard to balance 01:51 < bridge> we had smth like that but it was hard to balance 05:28 < bridge> so the practice bug ranks are still not cleared 05:28 < bridge> and saves 06:30 < bridge> morning 06:38 < bridge> gm 06:38 < bridge> any ec2 users 06:39 < bridge> can't figure it out, i have set inbound rules in a security group to allow all tcp traffic 06:39 < bridge> the firewall is disabled on the vm 06:39 < bridge> can connect on ssh but nothing else 07:55 < bridge> i got it 08:16 < ws-client> @learath2 well i gave you something 08:17 < bridge> Chiller how is your AB looking btw 08:19 < ws-client> good 08:19 < ws-client> ish 08:19 < ws-client> bad 08:19 < ws-client> depends 08:19 < ws-client> cant comment 08:23 < ws-client> @teero where is your C physics lib :c 08:33 < ws-client> im scared of multi threading is there something in ddnet that makes spawning and reusing a worker thread safe? Should I just use jobs? Feels like the wrong tool. I want some worker that will constantly have to compute something for me and return me a value. 08:33 < ws-client> its text file database :D so i feel like the code could look similar to the sql workers but its soooooooo complicated all the mutexes and atomic lock semaphores wtafak 08:36 < bridge> what does QuadOffset do when rendering pickups etc? 08:39 < bridge> @kebscs this is what happens when changing order with a mouse drag 08:39 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395656173843255397/ddnet_editor.mp4?ex=687b3d99&is=6879ec19&hm=9266069262b7b3c46737990a0d6d84a1507954b823499b0db10c5d739c006f73& 08:43 < bridge> changing the order by dragging pushes other layers around the one you are changing 08:44 < bridge> but undo swaps the index you moved with the index you moved the layer to 08:45 < bridge> can you reproduce this? 08:46 < bridge> you can just make a have a regular thread with some mutxes, maybe someone else will suggest something better idk 08:46 < bridge> I dont think you should use a job for that tho 08:46 < ws-client> REGULAR THREAD? 08:47 < bridge> Ddnet probably does it somewhere but idk if its reusable 08:47 < bridge> Yeah? 08:47 < ws-client> why do i pay for a ddnet game engine license if i have to write my own multi threading code??? 08:47 < ws-client> sounds dangerous bro 08:48 < bridge> Idk Im not aware that ddnet has what you want but I might be forgetting 08:48 < bridge> I think it only had the job queue 08:48 < ws-client> last time i touched threads was in 2008. It crashed my code so hard it took down the entire housing market 08:59 < bridge> :kek: 08:59 < bridge> @teero777 08:59 < bridge> (you fucked up pinging again chiller) 09:06 < ws-client> thanks 09:06 < ws-client> teeros gh name is 111 higher than his discord name??? 09:06 < ws-client> thats trol af 09:08 < ws-client> So if i want to check my workers state if its running or done i just do ``std::atomic m_State`` and can happily read and write to it from all threads? no risk what so ever? 09:15 < bridge> If you only care about that one int being possibly accessed simultaneously then I suppose yes 09:26 < bridge> @louis.place you can make this a draft PR directly (from the GH UI), so that others can't merge by accident. But I think the button is a bit hidden. Let me know if you're ready, I'd love to give you a review :owo: idea is brilliant 09:29 < bridge> @louis.place you can make this a draft PR directly (from the GH UI), so that others can't merge by accident. But I think the button is a bit hidden, [here](). Let me know if you're ready, I'd love to give you a review :owo: idea is brilliant 09:41 < bridge> Sadge 09:45 < bridge> @jupeyy_keks good morning! 09:45 < bridge> 09:45 < bridge> If we talk code-style in ddnet-rs, my impl in ddnet-rs#128 uses magic numbers all over the code and I thought about just adding consts for them, but it gets quite.. ugly. So what is your preference here? 09:45 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395672828229062668/image-11.png?ex=687b4d1c&is=6879fb9c&hm=260cf64e79e3d8c17dcc44eaa6b43e8832cdd00410a7cefe9645794d34f3368a& 09:45 < bridge> https://github.com/ddnet/ddnet-rs/pull/128 09:57 < bridge> Const can't hurt, right? 09:58 < bridge> Because then when you need the same value in more than one place you have to use a const anyways 09:59 < bridge> Yeah it's more about the fact that a single tab uses over 30 consts and they're not really reusable apart from margin values between components 09:59 < bridge> Because margin and spacing values are the only ones that should! Stay consistent between components :justatest: 10:00 < bridge> Because margin and spacing values are the only ones that should! Stay consistent :justatest: 10:01 < bridge> But what is the issue with consts? 10:02 < bridge> Even if there are 30 of them? 10:03 < bridge> I could use an enum struct 10:09 < bridge> Idk how ddnet-rs is set up, but it might be polluting the "global namespace" 10:14 < bridge> I've nothing against it, but if these constants are only used in a single function simply put then into the function body 10:15 < bridge> Cool didn't know the bot can do that xd 10:15 < bridge> just cleaned my pc 10:15 < bridge> so much dust 10:15 < bridge> but just in case in summer cuz heat 10:19 < bridge> 😅 10:19 < bridge> Got it :owo: 10:23 < bridge> I should do that as well, did you just use compressed air? 10:29 < bridge> i have a air gun 10:29 < bridge> with non static brushes 10:29 < bridge> 25€ 10:30 < bridge> with anti static electricity brushes 10:30 < bridge> I have an air compressor, which I can use on my car tires. Works for PC as well 10:32 < bridge> I have an air compressor, which I can use on my car tires. Works for PC as well, ~80€, but do recommend if you have a car or other compressed air usages 10:48 < bridge> I feel like I've seen prediction of other tees to clip through wall often these days, which wasn't the case before. Also I think when tees have high vertical speed, the client could have predict your position correctly. But instead your tee is flickering and in spec it's slowly falling down. 11:11 < bridge> This might be an effect of how intratick works on the client, it interpolates between two ticks but for some reason it has no upper bound when the server doesn't send snapshots which makes tees continue their trajectory forever. So it could be caused by packet loss of downloading snapshots or some other client bug 11:11 < bridge> That might be an effect of how intratick works on the client, it interpolates between two ticks but for some reason it has no upper bound when the server doesn't send snapshots which makes tees continue their trajectory forever. So it could be caused by packet loss of downloading snapshots or some other client bug 11:12 < bridge> When youre in spec the client doesn't do any prediction btw 11:13 < bridge> Although that does sound weird if its also happening not in spec 11:16 < bridge> Opinions on the other design idea in https://github.com/ddnet/ddnet/pull/10521#issuecomment-3084553517 ? 11:18 < bridge> you mean footer/header vs integrated? 11:18 < bridge> yes 11:18 < bridge> love this PR, i am in the end fine with both 11:18 < bridge> also whether we should increase the height of the graphs again because the footer/header reduce it 11:19 < bridge> love this PR, i am in the end fine with both, even tho I prefer the header/footer one 11:20 < bridge> do you see much value in the exact height of the graphs? For all of them imo it's more important to see dips and heights 11:23 < bridge> Not really, the relative differences are more relevant I think. If it was easier, it would have also been cool to add performance graphs for CPU or memory 11:24 < bridge> I wonder how you did the valramp graphs, are they always there or is there some dynamic graph selecting? 11:24 < bridge> They are enabled with `dbg_tuning 2` 11:24 < bridge> Used to be `dbg_tuning 1`, which only shows changed tunes now 11:25 < bridge> The values for the graphs are generated based on the tuning values 11:26 < bridge> interesting, because the valramp graphs should only be depending on the velramp values, would be interesting to have a marker where you are currently at 11:30 < bridge> just noticed i had some kernel options related to amd gpio disabled 11:31 < bridge> i wonder what that is 11:33 < bridge> that sounds to me a prediction bug tho 11:34 < bridge> The client just displays the server snapshots it received when youre in spec so it doesn't have any antiping jitters 11:34 < bridge> The client just displays the server snapshots it receives when youre in spec so it doesn't have any antiping jitters 11:34 < bridge> Its intended to work that way 11:34 < bridge> I mean it shouldn't jitter even when prediction is on 11:35 < bridge> since no input is actually changed no? 11:35 < bridge> or anything only the server knows 11:35 < bridge> or is there anything only the server knows 12:04 < bridge> Idk about arrows instead of min/max but other than that it's nice 13:22 < bridge> if theres magtic numbers like that, you can either comment them where they are `// x = points_props_width * y` or put them next to their usage or at the top of the code block 15:01 < bridge> Okay, turns out creating a crashdump, not trivial, I'm considering just biting the bullet and bringing in crashpad 15:03 < bridge> However, the nix crate in Rust, excellent to use, very nicely wrapped 15:03 < bridge> @learath2: ddnet related? 15:03 < bridge> Keep in mind: offtopic = ban 15:04 < bridge> Yeah, I was looking into improving our crash reporting situation. Turns out it's quite complicated to extract a useful dump 15:05 < bridge> I have it unwinding the stack and recording the instruction pointer for now, but you need much more, without some memory and context it's all useless 15:05 < bridge> Dump to file? 15:05 < bridge> On assert? 15:06 < bridge> Just a coredump is an option, but coredumps are big, not great for uploading 15:06 < bridge> Pls send crashes to ddnet.org 15:06 < bridge> Just a coredump is an option, but coredumps are big, not great for uploading and storing 15:06 < bridge> Nix mentioned 15:07 < bridge> not that nix 15:07 < bridge> https://tenor.com/view/nix-nixos-nix-my-beloved-linux-gif-16687923917201934033 15:07 < bridge> https://tenor.com/view/pucnh-goku-fight-angry-one-punch-man-gif-9678179172410065416 15:07 < bridge> I'll keep messing around with it though, it is kinda enjoyable learning about these things, how to unwind a stack and stuff 15:29 < bridge> coredumps are pretty well conpressible 15:30 < bridge> compressible* 15:33 < bridge> especially as a fair amount of the data should be quite similar between different coredumps 15:33 < bridge> Not bad, but around 80M for each crash report still might be beyond what is acceptable 15:33 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395760308798226646/image.png?ex=687b9e95&is=687a4d15&hm=4d7defc65b3323ae0c8eeac9538e84a9bd71c97a58e3a1ffbd8887062faff268& 15:38 < bridge> For Windows there's https://github.com/jrfonseca/drmingw which we already use for our crash dumps. It can also be used as a wrapper to launch another application I think. 15:39 < bridge> Or is this only about internal crash handling of servers? 15:39 < bridge> also let's not kid ourselves, certain people will try to spam this 15:43 < bridge> blog post inc 15:55 < bridge> this random minimized coredump generator I found generates 42K cores 😛 15:55 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395766050607468725/image.png?ex=687ba3ee&is=687a526e&hm=60b122b544c3d8150bedb17e16804bb8786beee1df6dbc864ae919a8d22102ed& 15:56 < bridge> What is included in the crash dumps we get? Are they windows minidumps? 15:57 < bridge> See https://discord.com/channels/252358080522747904/745926398140612678/1395717289155891250 for example. They are not minidumps. 15:59 < bridge> Yeah, are these really useful? It's a poorly symbolized stacktrace 15:59 < bridge> You need the symbols from https://ddnet.org/downloads/symbols/ for that particular git hash and release type 16:03 < bridge> Hm, I guess it's better than nothing, at least we get a trace 16:04 < bridge> From the DLL name a launcher could also automatically determine if it crashed in the graphics driver and suggest changing graphics settings 16:14 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395770754179661914/image.png?ex=687ba84f&is=687a56cf&hm=86ebc1e52df0cdd81bce1f7564adb420a2bc1ad98dd7bb70a505cf34c3cf65ea& 16:14 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395770754557284522/image.png?ex=687ba850&is=687a56d0&hm=08afd3ff06866c9c154163e89334f360bead4dd8f0b87c9fb1248396e5ac8abb& 16:14 < bridge> pog optimizations 17:07 < bridge> idk if heres the place but is something with the cheating and tas problem going to change or is it doomed? 17:09 < bridge> doomed 17:13 < bridge> :JoeSuperAngry: 17:13 < bridge> :ahhh: 17:14 < bridge> couldnt we add some type of anticheat or a anti tas squad that deletes tas on popular maps:thumbs_up_joe: . cheaters can do whatever and get away with it atm. just restart router or get a vpn done then you back :C 17:16 < bridge> I understand this is a opencource. just wondering if anything can be done :D 17:18 < bridge> admins do have anticheat and replays with teehistorian 17:19 < bridge> will they do anything+ 17:19 < bridge> as u see 17:20 < bridge> hope they will 17:20 < bridge> really enjoy this game and speedrunning it 17:21 < bridge> @axel628: Are you trashcleaners? 17:21 < bridge> could be xD 17:21 < bridge> I like your mult bot bdw <3 17:21 < bridge> So mysterious 17:22 < bridge> could you add !topspree? 17:22 < bridge> 🙏 17:22 < bridge> soon(tm) 17:26 < bridge> and go more left for kill to count is annoying to get 20 times even if you muted the bot🙏 17:28 < ws-client> that is so true sorry for spamming will fix @PRO_Gaming1921 17:28 < ws-client> spawnblocker ban btw <:banhammer:392813948858269696> @PRO_Gaming1921 17:28 < bridge> :ahhh: 17:35 < bridge> it isnt spawnwblocking. whole playground should be counted as a kill 17:35 < bridge> most of people block far < as do you :joobisaythatagain: 17:35 < bridge> you kill people there I see 17:37 < bridge> Nah bro entry blocking is cring af 17:39 < bridge> @axel628 BAD PIPOTAM 17:39 < bridge> getting 20 spam is cringe af =((( 17:39 < bridge> https://tenor.com/view/pipotam-baby-hippo-hippo-wash-shower-gif-3406421565683572068 17:39 < bridge> https://tenor.com/view/mario-super-mario-bros-depression-sad-mario-and-luigi-gif-14815881022592239601 17:40 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1395792337774575647/image.png?ex=687bbc69&is=687a6ae9&hm=acc708214698c0d106346acfcae103c8ebccaaa0b0f964219f84ff30c6a202dd& 17:40 < bridge> < filters nubs 17:40 < bridge> needed 17:40 < bridge> there all block 17:44 < bridge> Spam Bad and I will fix yes 17:44 < bridge> tyty! 18:13 < bridge> have u ever taked a look at the teehistorian data 18:13 < bridge> i wanna try some stuff when im not busy 18:13 < bridge> have u ever took a look at the teehistorian data 18:13 < bridge> have u ever taken a look at the teehistorian data 18:14 < bridge> the format is lowkey cooked though xd 18:30 < bridge> no 😂 18:30 < bridge> but twgame has stuff for this 18:31 < bridge> its in rust unfortunately 18:36 < bridge> yeah i am not building teehistorian data scripts in rust 18:36 < bridge> gonna have to make some python shit ig 18:56 < bridge> i told Cammodude that we can use my easy antibot temporarily but he didnt like it 18:56 < bridge> cammodude didnt make bot? 18:58 < bridge> aah alr. but you working on it? 18:59 < bridge> i done it 18:59 < bridge> cammodude works on real antibot 18:59 < bridge> 🫡 19:22 < bridge> https://github.com/tracel-ai/burn 19:22 < bridge> https://burn.dev/blog/sota-multiplatform-matmul/ 19:29 < bridge> swappable backends is pretty cool 19:52 < bridge> @kebscs yeah i was unsure how to handle gun 19:52 < bridge> technically every weapon has a unique niche 19:52 < bridge> for example hammer goes thru walls, nade has arcs, laser can bounce 19:52 < bridge> pistol seems a bit too easy though, maybe i just remove it? 19:53 < bridge> as in bullets still hit targets but wont activate 19:53 < bridge> it can still be used in telegun parts. also it'd ensure ppl don't map annoying parts where you have to shoot something off screen 20:24 < bridge> gosh, cyrillic broke every single regex in my script just now 20:24 < bridge> :angy: 20:24 < bridge> spammed the shit out of ddnet.org, ggwp 20:34 < bridge> Ddosser 21:01 < bridge> it’s easy 21:01 < bridge> not as easy as my precious python 21:07 < bridge> https://youtu.be/DNBKdU6XrLY?si=p5ZiWI3WVeqPFCEk 21:07 < bridge> what do you guys think about the sand pond saga? 21:17 < bridge> accounts would fix it 21:19 < bridge> yea but gun is too op 21:19 < bridge> cant do parts like on your example vid 21:19 < bridge> or shooting straight with laser 21:19 < bridge> or parts where you have to find the laser to shoot the target 21:23 < bridge> So why is it taking so long??? 21:23 < bridge> Why would you even design a game with such a flaw in the first place 21:23 < bridge> Ofc accounts are needed 21:29 < bridge> it was designed as a fun little mod for Teeworlds, not as a game 21:30 < bridge> Even kog has accounts 21:30 < bridge> It's so simple obviously 21:30 < bridge> Just make accounts 21:30 < bridge> And be happy 21:31 < bridge> but simple boring accounts are boring and simple 21:31 < bridge> we're open source developers here 21:31 < bridge> no point in making a system that isn't novel and decentralized and that might get finished in a reasonable amount of time 21:33 < bridge> You could probably also cryptographically sign finishes or something 21:33 < bridge> Idk 21:33 < bridge> Or use blockchain 21:33 < bridge> I think we need Blockchain in DDNet 21:35 < bridge> Actually it's kinda funny how steam items are like NFTs 21:35 < bridge> But just without crypto stuff 21:35 < bridge> And somehow people don't mind it 21:37 < bridge> In practice crypto seems to be only useful for illegal stuff 21:37 < bridge> ik 21:37 < bridge> Why 21:37 < bridge> but it seems like it wont be a possibility 21:38 < bridge> Someone was working on an account system 21:38 < bridge> And i wonder what's the progress 21:38 < bridge>