01:28 < bridge> Is a score of 0 even a possibility, we use this score thing in such a bizarre way 😄 04:45 < bridge> lmao, some people thought we made a inferior IME into ddnet and they are now asking me how to turn it off cuz they thought it wasn't smart compared to their system IME 04:52 < bridge> And is that true? 04:53 < bridge> I'd have thought it's exactly the same from the suggestions 04:59 < bridge> what suggestions 05:13 < bridge> oh candidates 05:13 < bridge> It should be the same as their IME 05:13 < bridge> We couldn't possibly built-in multi-language IMEs directly into ddnet 05:14 < bridge> It's probably just the look that is throwing people off 07:27 < bridge> Ok, then they will get used to it^^ 07:31 < bridge> that's what i thought 07:31 < bridge> But some people just can't understand they are using their exact IME 07:31 < bridge> They insist we made a dumber version of chinese input lol 07:46 < bridge> it is, and before it didnt show in info, but i made it show 08:15 < bridge> make sure u test this properly 08:15 < bridge> i wont fix i t 08:16 < bridge> point based has to work too 08:16 < bridge> it looks good on a quick look 08:26 < bridge> @Robyt3 sir, I forgot about the people who uses arrow key to select candidates 08:26 < bridge> :justatest: 08:31 < bridge> I'd also do that 😂 08:32 < bridge> you'd use arrow keys or forget them 08:32 < bridge> Use arrow key 08:32 < bridge> Oooo 08:35 < bridge> Do you not do so when using Google or coding ? ^^ 08:47 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111546121328799765/61f2fe1c3a542848.jpg 08:47 < bridge> I don't need to select candidates on my own ime 08:47 < bridge> But if i do i just use numbers 08:48 < bridge> Hi fred 08:50 < bridge> do we have a config to turn off candidate window 08:50 < bridge> People are asking for it 08:50 < bridge> Wtf. Even the fps are overlapped 08:51 < bridge> Idk what happened, just forwarding bug reports 08:51 < bridge> It's ok. Just never seen such a bug ^^ 08:53 < bridge> Oh wait i think arrow key works 08:53 < bridge> It's just not what they used to 08:54 < bridge> Maybe we should collect all these issues 08:55 < bridge> Is any of them at least happy to be able to play in Fullscreen 08:55 < bridge> I dont think the arrow key can be adjusted 08:55 < bridge> It's entirely handled by ime 08:55 < bridge> Playing windowed under windows must suck hard ^^ 08:55 < bridge> Probably but the happy ones don't talk 08:55 < bridge> Xd 08:55 < bridge> That's normal i think 08:59 < bridge> I think the background was no clear 08:59 < bridge> Without anything drawn on it 09:02 < bridge> Maybe driver bug? 09:03 < bridge> Generally looks really buggy to me 09:15 < bridge> why are there so many driver bug that we keep getting 09:15 < bridge> :justatest: 09:16 < bridge> Because it's cheaper to produce shitty drivers 09:17 < bridge> ~~Let's replace our renderer with Godot's graphic server~~ 09:17 < bridge> I wonder how much work around they have 09:17 < bridge> And black listing 09:18 < bridge> I randomly scanned throw some steam games a while ago. We are definitely not the only ones with these problems 09:18 < bridge> Through 09:19 < bridge> Is it generally better using latest drivers on windows? 09:19 < bridge> Most of the time yes. But not always 09:19 < bridge> I still don't quite understand how we manage to hit so many while rendering a 2d game 😄 09:19 < bridge> Should i tell them to update drivers? 09:19 < bridge> But windows often has drivers shortly after the hardware released xD 09:19 < bridge> They often suck 09:20 < bridge> Are there so many reports like this? 09:20 < bridge> Sounds weird to me as well but I'm not the graphic guy so idk 09:20 < bridge> Just one 09:20 < bridge> Doubt it. So what us the issue anyway? They don't like that we render the candidates? 09:20 < bridge> I just told them to reset their custom background if they can 09:20 < bridge> I guess u could ask if it happened after alt tab or smth 09:20 < bridge> Some people don't like it 09:22 < bridge> Hm, I wonder what we are missing. We can certainly add q config for it but maybe ask them what the issue is? 09:22 < bridge> Or ask them to show their nornal ime completion in other applications.m 09:23 < bridge> I can imagine there being some extra information about the candidate in the window maybe but that's about it 09:23 < bridge> We are missing nothing 09:23 < bridge> They want their ime "skins" back 09:24 < bridge> For the windowed users 09:24 < bridge> I don't feel like it's many. It's just that every person having trouble reports it. If all ppl without problem would say smth u'd get a feeling for it.. most annoying is if discord overlay or similar destroys the game for no reason ^^ 09:24 < bridge> From 0 to 100 how sure are you? 😛 09:24 < bridge> 20 09:24 < bridge> Users are rarely completely wrong when reporting regressions in usability 09:25 < bridge> We can add input state where it shows which language you are currently typing 09:25 < bridge> We can add ime state where it shows which language you are currently typing 09:25 < bridge> But lots of third party ime misreport that 09:25 < bridge> It's a feature paradigm shift 09:26 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111556019273859152/-6c18cab31aa38140.jpg 09:26 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111556019487772673/-471c170505cecf64.jpg 09:26 < bridge> Some people just want their waifu back 09:27 < bridge> There is ofc bias there but still. Even the variety of bugs we manage to hit is impressive. I'd think we were using only the tiniest part of the driver which probably gets tested a lot in QA 09:28 < bridge> It's not like we hit some insane codepath while using a vk extension that is only standardized by one vendor 09:28 < bridge> Yep. I'm also impressed. Especially about stuff like hibernate or alt tab, which often cause trouble 09:29 < bridge> Okay. Now I understand. The config is of utmost importance, this is indeed a regression 09:30 < bridge> but we can't bring the window back with unmodified SDL 09:30 < bridge> we have to patch SDL more to get it togglable 09:31 < bridge> Huuh? What needs to change in sdl? 09:31 < bridge> we shut-off sdl's input method handling before to stop the window capturing the input method state and goes into headless ime mode 09:32 < bridge> we shut-off sdl's input method handling before to stop the game window from capturing the input method state and goes into headless ime mode 09:32 < bridge> that was just a hack to let windows do it's normal IME things 09:32 < bridge> we can bring back the hack and make it togglable some how 09:34 < bridge> but I'm pretty sure half of the users are experiencing flashing screen during IME inputs before. the current solution works for most people and just make it look ugly for 5% of the people who has a older dual GPUs and plays in windowed mode 09:34 < bridge> the older hack only works if they have an older style dual GPUs setup on laptops 09:35 < bridge> Hm, my knowledge of the IME mess has grown hazy but I thought IMEs didn't even attempt to render if sdl text input wasn't active 09:36 < bridge> it was active always 09:36 < bridge> with the hack 09:36 < bridge> Wouldn't it work for everyone playing windowed or borderless? I thought the dual gpu thing was only for fullscreen to not flash 09:36 < bridge> because sdl didn't tell windows to stop listening to text input 09:36 < bridge> And how did we stop it from triggering on, say movement keys? 09:37 < bridge> nothing 09:37 < bridge> user turns off their IME 09:37 < bridge> And they liked that workflow better :confusion: 09:37 < bridge> i think they just think it is ugly 09:37 < bridge> i'm 95% sure 09:37 < bridge> on this one 09:38 < bridge> Ok does winapi provide a `GetCandidateWindowWaifu_Ex`? 09:38 < bridge> i would've gone that route in teeworlds 0.7 if i could 09:39 < bridge> Or we can commission Voxel to draw us a ddnet waifu to put on our candidate window 09:39 < bridge> Xd 09:39 < bridge> :justatest: non-weeb would be furious 09:39 < bridge> I can also draw waifus now 09:39 < bridge> i mean we could make the candidate window skinable 09:39 < bridge> It's easy to config gate 😄 09:39 < bridge> but is that what we REALLY need? 09:40 < bridge> I'd happily take on the skin config development if we really need that 09:40 < bridge> Idk, I would also be furious if you took away my waifu and cute rabbit thing 09:40 < bridge> at least a horizontal mode would be nice, since windows does not report that, we could just make it a skinning feature 09:41 < bridge> If possible please poll around a bit, make sure the issue is indeed only a looks issue. Then we can focus in on a fix to make the 5% happy-ish-er 09:42 < bridge> are we sure we know they know what they want tho 09:42 < bridge> I mean they should have an idea why they don't like it, no? People who genuinely only hate the looks will usually just say so 09:43 < bridge> they say they don't like the look and want their own ime window back 09:44 < bridge> the four people complaining are pretty clear about that 09:45 < bridge> some do they they forgot how to type with a different looking candidate window 09:45 < bridge> Okay so wait, don't we now toggle the sdl input mode properly? Why isn't their IME triggering? How are we blocking that? 09:46 < bridge> this was one of them, but i think the feature parity is there. they're just not familiar with the interafce 09:46 < bridge> sdl tell windows to use headless ime mode 09:46 < bridge> Is that hardcoded in SDL? 09:46 < bridge> i think so 09:47 < bridge> let me pull out sdl code 09:47 < bridge> Wait is it SDL providing us the candidates? 09:47 < bridge> yes 09:47 < bridge> well no 09:47 < bridge> Ah cute, this wasn't a thing back when I last checked 09:47 < bridge> but we are getting the window message through SDL 09:47 < bridge> iirc 09:47 < bridge> it's been a while 09:48 < bridge> I can't imagine a lot of games draw their own IME, maybe there is a way to make sdl not do that? 09:48 < bridge> there was a candidate window thing in SDL that is just empty. but i think we are getting WM_MSG or something using SDL's event 09:48 < bridge> a lot of games don't care about IME 09:49 < bridge> the ones that care (i.e. source engine) draws their own 09:49 < bridge> unity does not do anything with candidate, fullscreen = no candidates 09:50 < bridge> I'm pretty sure I've done the research when I was working on it years ago. :kek: 09:51 < bridge> Eh. Just take a quick look, see if you can either override sdl asking for headless mode (maybe by a call to winapi on our side) or better tell sdl not to do that. If neither looks feasible I guess people will have to live without their waifu for a sec 09:51 < bridge> i'm doing that 09:51 < bridge> at work 09:51 < bridge> with my boss looking at me 09:52 < bridge> Maybe could be worth asking the players that complain that if a horizontal mode would help 09:52 < bridge> This is capital theft. Please stop it 09:53 < bridge> I think sdl recently changed it so candidates are send 09:53 < bridge> That's why we updated sdl 09:53 < bridge> I should try out kde ime again. But as always in Linux. I can use it in Fullscreen without problems 09:54 < bridge> So it's not needed there xd 09:54 < bridge> I bet they don't support linux properly anyway 09:54 < bridge> SDL was mainly build with Linux in mind 09:54 < bridge> https://github.com/ddnet/ddnet/blob/a5332f0b1177406610c819636850525549186d60/src/engine/client/input.cpp#LL789C28-L789C30 09:54 < bridge> It's sponsored by valve xd 09:54 < bridge> we are grabbing windows event 09:54 < bridge> Yeah but none of the linux imes expose their candidates ever 09:55 < bridge> i think linux IMEs are just a bunch of different user space programs doing hacks 09:55 < bridge> Mh ok, sucks 09:56 < bridge> True😂 09:56 < bridge> there's no way we could know what the IME is doing 09:56 < bridge> Yeah so sdl basically just added that headless mode thing and allowed wm events to passthrough? 09:56 < bridge> I always used the smiley ime popup window xd 09:56 < bridge> they didn't add it, it was there since SDL2 09:57 < bridge> we removed it in our custom SDL before 09:57 < bridge> they did also tell windows when to stop activating text input / ime so that's good. but we did also removed that before 09:58 < bridge> That is not how I remember that piece of code at all, but if we are being honest I spent more time on the macOS part of that code anyway so maybe I just missed it 09:59 < bridge> Ok so we only need the new sdl for passthrough wm events? 09:59 < bridge> macOS can draw candidate window on top of fullscreen application tho 09:59 < bridge> i actually don't know what we're needed for the new SDL. robyt3 knows more about that. it was always working in teeworlds, so I didn't know if they are using a newer SDL 10:01 < bridge> I think candidate was working fine in macos. MacOS only needs a drawing position other than the StartInput and StopInput thing which we are providing through SDL. 10:01 < bridge> I think candidate was working fine in macos. MacOS only needs a drawing position which we are providing through SDL, other than the StartInput and StopInput thing 10:01 < bridge> I think candidate was working fine for teeworlds 0.7 in MacOS. MacOS only needs a drawing position which we are providing through SDL, other than the StartInput and StopInput thing 10:01 < bridge> sorry IRC 10:02 < bridge> macOS had a massive issue with sdl not allowing access to the underlying fake textinput container they create to trigger the candidate window 10:02 < ws-client> SDL version-specific issues 2.26.5, 2.24.2, 2.0.22: IME candidates work. But there are minor bugs when moving the composition cursor. 2.0.18, 2.0.20: IME candidates work. 2.0.16 (our current version): IME candidates cannot be determined with Windows API. Windows tries to draw the composition window like before, so this does not work in fullscreen mode. 2.0.8 (upstream 0.7): IME candidates work. But this SDL 10:02 < ws-client> version is too old for us. 10:02 < ws-client> from #6530 10:02 < chillerbot1> https://github.com/ddnet/ddnet/issues/6 10:02 < ws-client> chillerdragon: ^ fix bot 10:02 < bridge> bad bot 10:03 < bridge> The textediting events you got refered to nonsense offsets for the small PressAndHoldIME 10:04 < bridge> 10:04 < bridge> I guess the question is what 2.0.16 was doing 10:04 < bridge> Though I wonder if robyt was referring to our patched version of 2.0.16 10:05 < ws-client> i mean whats the problem, lets just use the current SDL version 10:05 < ws-client> what would it help to downgrade again 10:06 < bridge> 2 concerns 10:06 < bridge> 1) weren't some people complaining of input lag? 10:06 < bridge> 2) we took away peoples waifus 10:06 < ws-client> 1.) is bcs of mouseold 10:06 < ws-client> we should simply remove it 10:06 < ws-client> 2.) we can still make it config 10:06 < bridge> Damn mouseold... 10:06 < ws-client> how does old version help here 10:07 < bridge> I was about to say that TODO comment looks wacky 10:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111566236241903626/image.png 10:07 < bridge> ah, damn it 10:07 < bridge> 2) Seems something changed after 2.0.16 which makes windows no longer trigger normal imes idk 10:08 < ws-client> but we do that on purpose dont we? 10:08 < ws-client> we probably dont set the input rect 10:08 < bridge> oh wait, is it possible robyt3 is unaware of our hacked IME patch 10:08 < bridge> so he just thought it was broken in our current version 10:08 < ws-client> @Robyt3 ? 10:08 < bridge> It is indeed possible 10:08 < ws-client> enlighten us 10:09 < bridge> because we did turn off all IME stuff in our custom SDL just to make IME show up in windowed mode 10:09 < bridge> that would be my fault 10:09 < bridge> Anyway, did you manage to check? Does sdl unconditionally enable the headless thing? 10:09 < bridge> oh 10:09 < bridge> still trying to find the code 10:09 < ws-client> i just dont understand why SDL prevents it, if they clearly have an API function to tell where to render the candidate list 10:10 < bridge> it was for multi-platform, macos need it 10:10 < ws-client> Note: If you want to use the system native IME window, try setting hint SDL_HINT_IME_SHOW_UI to 1, otherwise this function won't give you any feedback. 10:10 < bridge> also they might have did the windows one before they decided to grab candidate themselves 10:10 < ws-client> its all only for macos? xd 10:11 < ws-client> https://wiki.libsdl.org/SDL2/SDL_SetTextInputRect 10:12 < ws-client> we have so many #ifdef WINDOWS.. why do we even use SDL in first place :joy: 10:12 < bridge> That hint sounds about what we want, but idk if they wired it up properly 10:16 < bridge> found it https://github.com/libsdl-org/SDL/blob/d4bc393efebdd45a95430397df106b69b1c26df3/src/video/windows/SDL_windowskeyboard.c#LL412C9-L412C31 10:16 < bridge> `return SDL_GetHintBoolean(SDL_HINT_IME_SHOW_UI, SDL_FALSE);` 10:16 < bridge> ok so it is a SDL_HINT 10:17 < bridge> I guess we can add a config for it if it does what we want 10:17 < bridge> it would probably require a restart after the config change 10:17 < bridge> but it might be possible 10:18 < bridge> I heard many people say that. did anyone make a benchmark of how long common things take? 10:19 < bridge> It's an ancient meme at this point. It's probably a couple microseconds faster on everything but this isn't the early 2000s anymore 10:20 < bridge> Maybe you can get it up to a couple millis if you are brave enough to enable systemwide O3 and LTO 10:20 < bridge> But a couple of microseconds that add up to each ither 10:20 < bridge> I once read around 5-10% more perf can be expected 10:20 < bridge> Sometimes also degrations 10:21 < bridge> I mean fedora enabled system wide lto 10:21 < bridge> And opt out when pkgs broke 10:21 < bridge> From systemwide o3 and lto? 10:22 < bridge> Or just one or the other? 10:22 < bridge> Can't remember, but smrh with cpu native etc march native 10:22 < bridge> the kernel also has a option to enable zen3 optimizations 10:22 < bridge> But can't be hard to find benchmark. But do u know Intel clear Linux? 10:22 < bridge> It's basically that 10:22 < bridge> https://wiki.gentoo.org/wiki/Ryzen 10:22 < bridge> And it often is a bit faster 10:23 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111570201775972352/image.png 10:24 < ws-client> ^ i think phoronix had an article about that once and results were mixed, sometimes faster sometimes slower 10:25 < ws-client> https://www.phoronix.com/news/Linux-5.19-O3-March-Native 10:25 < ws-client> ok was about kernel 10:26 < ws-client> https://www.phoronix.com/benchmark/result/core_i5_12600k_kernel_build_optimizations/35ca2579382e.svgz 10:26 < ws-client> ufs xd 10:26 < ws-client> https://www.phoronix.com/review/gcc-10900k-compiler 10:26 < ws-client> side doesnt load for me 10:26 < ws-client> but that sounds promising 10:27 < bridge> site doesn't load for me either 10:27 < bridge> but it's not quite what I was aiming for 10:27 < bridge> I wanted to see whether tools on gentoo are typically faster than tools on arch 10:28 < ws-client> https://web.archive.org/web/20221110002751/https://www.phoronix.com/review/gcc-10900k-compiler/2 10:28 < ws-client> tools = build tools? 10:28 < bridge> I wasn't aware of this, but I think I downloaded all 2.0.18 directly from github, so it should be without our own changes, unless they were also in upstream SDL 10:29 < bridge> Mh, I can only theorize, we can maybe test. But I imagine every single tool is a couple millis or micros faster for all tasks 10:30 < ws-client> i mean if benchmarks are faster, then tools are too xdd. except its heavily IO dependent or smth 10:31 < ws-client> ok u could measure the startup time of the tools, but lto is probably unbeatable here too, esp if u static link as much as possible 10:32 < ws-client> fyi: jupsti im quickly restarting the backend 10:33 < bridge> > i mean if benchmarks are faster, then tools are too xdd. except its heavily IO dependent or smth 10:33 < bridge> only if you assume that the distros ship everything the same except for the compilation flags 10:33 < bridge> perhaps gentoo is disabling the systemd integration of cat, which makes things go slower for whatever reason 😄 10:34 < ws-client> . 10:34 < bridge> https://jvns.ca/blog/2023/05/25/new-playground--memory-spy/ 10:34 < bridge> Ah well that's true ^^ 10:34 < bridge> systemd-catd 10:34 < bridge> @deen did you try the IP selection thing for cn or did it fixed itself 10:34 < bridge> Mention systemd, summon Learath 10:35 < bridge> I mean there **is** systemd-cat 10:35 < bridge> but it's doing something that you're probably not unhappy about ^^ 10:35 < bridge> Please tell me it's not replacing coreutils cat 10:35 < bridge> correct 10:35 < bridge> I'm thinking if we could build-in a IP tester in engine, could it also fix clients not be able to get serverlist without we having to setup a hk serverlist without CDN 10:35 < bridge> it's not doing anything related to coreutils `cat` 10:36 < ws-client> jupsti how are you always online? dont you have a job or something? 10:36 < bridge> what's a IP tester? 10:36 < bridge> what's an IP tester? 10:36 < ws-client> chiller i could ask the same about you 10:36 < ws-client> :justatest: 10:36 < ws-client> i am at work rn ._. 10:37 < bridge> Randomly testing different cloudflare ips and force curl to use that instead of rely on DNS 10:37 < bridge> In china 10:37 < bridge> Randomly testing different cloudflare ips and force curl to use one in good condition instead of rely on DNS 10:38 < bridge> ah 10:38 < bridge> DNS can not tell whether china is corporating but we might be able to test it ourselves 10:38 < bridge> I would be surprised if everyone isn't shipping the exact same e.g. coreutils barring maybe a version difference 10:38 < bridge> our current setup isn't bad for that either, no? 10:39 < bridge> I don't think everyone is shipping it in the same configuration 10:39 < bridge> Occasionally some one has to wait for half and hour before they can grab a serverlist when the list server is down 10:39 < bridge> And there were some map downloading issues 10:40 < bridge> because we can't keep up our serverlist server? 10:40 < bridge> Occasionally some one has to wait for half and hour before they can grab a serverlist when the hk list server is down 10:40 < bridge> I think maps are downloaded from somewhere else in china, no? 10:40 < bridge> because we can't keep our serverlist server up? 10:41 < bridge> yes but I'm not sure if which way they were going with only screenshot reports 10:41 < bridge> Also might cut some cost for cn if cloudflare could just work better 10:42 < bridge> Just a thought tho, might be worth a experiment 10:43 < bridge> Btw occasionally = roughly one per week 10:44 < bridge> why is the list server down so often? 😮 10:44 < bridge> One report per week 10:44 < bridge> ah 10:44 < bridge> The server was ddosed twice for the last month i think 10:44 < bridge> is there something like object storage in china? 10:45 < bridge> Yes map were using it 10:45 < bridge> can you link it (in DM if it's sensitive)? 10:45 < bridge> But we couldn't use master2.ddnet.org for that 10:45 < bridge> why not? upload a servers.json every 5 seconds 10:46 < bridge> or is that too expensive? 10:46 < bridge> we have to claim ddnet.org and register it with gov to bind cdn or obejct storage to a custom domain 10:46 < bridge> If we can just use other domains then it should be fine 10:47 < bridge> hmm, using other domains doesn't seem so trivial 10:47 < bridge> Ye i just don't think it looks good for ddnet tho 10:48 < bridge> Que someone exposing ddnet is phonehoming data to china 10:49 < bridge> A target registered with chinese gov no less 10:49 < bridge> so in order to use master3.ddnet.org in china with a big CDN, we'd need to do some stuff we're not willing to do? 10:49 < bridge> ye 10:49 < bridge> If it's data premier Xi demands, data we'll send :peepoSalute: 10:50 < bridge> we could use teeworlds.cn however we want cuz we did registered it 10:50 < bridge> Why not make master3 a CNAME then 10:50 < bridge> Then u can use a domain 10:50 < bridge> how does CNAME interact with HTTPS? 10:50 < bridge> probably still needs a certificate for the original domain 10:50 < bridge> ? 10:50 < bridge> which won't be possible in china with big CDNS 10:50 < bridge> Actually no idea 10:51 < bridge> Just use same certificate for both 10:51 < bridge> can we upload our own certificate to the cloud vendor, @TsFreddie? 10:51 < bridge> this would solve the problem and we wouldn't even need a CNAME 10:52 < bridge> at what point does the CDN require you to have control of the domain in china? 10:52 < bridge> let me check 10:52 < ws-client> jupsti xxdddd my cheap work laptop attached to a 2560x1440 screen on gnome reaches the same fps as my 4090 desktop at home xd (400fps on average all maps) 10:53 < ws-client> blame nvidia for bad interaction with gnome, but also blame gnome 10:53 < ws-client> bcs in openbox it works :D 10:53 < ws-client> good meme 10:53 < ws-client> yes 10:53 < ws-client> im so gonna switch away from gnome 10:53 < ws-client> epic gamer moment 10:56 < bridge> zzz 10:56 < bridge> try xfce 10:56 < ws-client> pls 10:56 < bridge> I feel a little uneasy trying to loop around the backs of the chinese govt 10:57 < bridge> where do u have to do this 10:57 < bridge> Chiller use i3 or bspwm. Convert to the tiled wm religion 10:57 < bridge> @heinrich5991 are you planning on uploading to storage using master1 btw 10:57 < ws-client> Lea: KDE also has tiling support :joy: 10:57 < bridge> or cdn a master2 10:58 < bridge> or either 10:59 < bridge> Fwiw I still think this just cache kinda approach to the master server is a bit lackluster except for a fallback. Servers in china won't be able to register through these fake masters 10:59 < bridge> cdn just doesn't want to use unregistered domain 10:59 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111579282796445756/image.png 10:59 < bridge> it doesn't matter what cert it uses 10:59 < ws-client> but if u register a random one 10:59 < bridge> i think if the final destination is tencent then they can just refuse to set it up 11:00 < bridge> oh i get it 11:01 < bridge> ye we probably can 11:02 < ws-client> i think first ill try dwm 11:03 < ws-client> omg my work internet is smooth af. And here i was unsure for months if my home net is laggy or if it is in my head 11:03 < ws-client> my home setup is broken in every way while my work setup is perfect gaming moment 11:03 < ws-client> i hate my life 11:05 < bridge> lol 11:05 < bridge> dwm is a lot of setup iirc 11:05 < bridge> moreso than i3 11:05 < bridge> i3 is good start 11:06 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1111581005720064091/image.png 11:06 < ws-client> from gnome to dwm xd 11:06 < bridge> how can we try tho 11:06 < bridge> what are u doing fred i missed it 11:06 < bridge> Create cert for master 3 and upload it there xd 11:07 < bridge> hmm, interesting idea 11:07 < bridge> @heinrich5991 do you have a rough idea how many request we are getting per day 11:07 < bridge> no 11:07 < bridge> ah 11:07 < bridge> let me check if cloudflare tells me something 11:08 < bridge> ah right, can't, because my password manager locked me out 11:08 < bridge> 😄 11:08 < bridge> we are looking at 0.2cny per 10k requests 11:09 < bridge> thats 0.026€ per 10k requests 11:09 < bridge> if it is less than 100 per month i can just eat it 11:10 < bridge> @TsFreddie I'd guess they'll just refuse to serve requests that have a `Host: ` that wasn't registered 11:10 < bridge> so CNAME wouldn't really be an option 11:10 < bridge> they can, but i don't know if they would yet 11:11 < ws-client> i guess they only want the cert to spy, dont they? 11:11 < bridge> well they probably would 11:11 < bridge> the cert is necessary to function as a CDN ^^ 11:11 < bridge> cloudflare also has a cert 11:11 < bridge> for ddnet.org 11:12 < bridge> that's 1000 online players all day averaging 6 requests per hour per player 11:13 < bridge> they can do what? 11:13 < bridge> they serve requests with a `Host: ` that wasn't registered in china like that? 11:14 < bridge> they can refuse 11:14 < bridge> :justatest: 11:14 < bridge> `Host:` header only has one domain right? 11:14 < ws-client>