#programming
1 messages · Page 94 of 1
idk either
‱
ye i cant finf it immediatly in win+v
but i found this weird percentage with too many 0
0/000, dividebyzeroerror
catastrohic failure
same
honeslt wehn you devide by 0 you should just start continuing with ±inf
inf check my beloved, multiplies the signs (I'm looking at you -0)
https://www.pexels.com/photo/brown-wood-surface-172289/ https://www.youtube.com/watch?v=CrQKA43aZrQ&t=231s https://free3d.com/3d-model/csgo-desert-eagle--93708.html I tried using this wood texture but instead it deleted the gun ?
uh what?
jerm
jer?
Oh man, it is funny that now companies are looking to ban AI art/content on their platform
good luck you got this
at least you got to the interview stage i never got that far 
not really a job interview more of a course interview
ah
got some notes on the second screen
firm handshake, be confident, and when they ask you to tell about yourslef dont explain why you're fit for the job, they actually want to hear a short description of you like yuor hobbies and siblings and stuff
e interview
thats what i learned in school a month ago at least, i have no clue if its acurate
but i got a physical one tomorrow
as well as an aptitude test
at least its through zoom and not teams
I do have a job interview coming up, pushed back to Friday so that's fun.
Ive done a few interviews now
and like 200+ applications now
1 offer and its IT, which is cool, but I'm already IT and I wanted to swap to eng or data
they want me to be 10 mins early but the meeting room isnt opened yet
Just join asap i guess
finished
interview for a course, did ok

j*b
based programmer wins against openai
in optimization coding tournament
Without ai tools

and not even with a slight margin, that is subtential
Ai is NOT a match to polish programmers
*Caffeinated Polish Programmer
Gemini 5
Gemini 2.51
Gemma 3 thinking when 😭

its probably the open-weight model they were planning to release but was pushed back
I like doing that
gius should i switch from nginx to OpenLiteSpeed?
Depends on your use cases. Do you have any specific need for the webserver to do?
If your issue is just configuration, then I can only say go for it
for small sites, like some privatebin selfhost for example
or just personal site
I think if the current setup works then why change?
i found openlitespeed kinda easier to use
because it has web dashboard
at the moment i use openlitespeed for phpmyadmin
Yeah, as I said then, if your issue is config and monitoring then go for it
For small site it doesn't matter much which one you choose
i care abt more ram/cpu usage
so i could save stuff for other projects aswell
Honestly I think it won't matter that much. You application or database will most likely have bigger memory footprint than the webserver.
abt that should i allocate more?/
Is that for one service or the whole thing? If it is for one service only, that is more than enough for small/personal site. If it is for the whole suite of application that you're going to install, then I suggest adding a bit more depending on how many services you'd like to run
i'm gonna use that for everything
that db
it runs mariadb
Oh, that's enough then. Its still count as one service technically (one database server). Maybe add one more CPU if you're experiencing lag spike
with that theres might be issue because idk if my cpu can handle it (its hp elitedesk used as homelab)
Yeah, I'm just suggesting it if you're experiencing lag-spike. Highly unlikely if you're the sole user and all the service installed is behaving properly
so far my setup looks like that:
vm1 = php sites (only, runs nginx)
vm2 = mariadb and mongodb (i didnt get it yet)
vm3 = docker and nodejs discord bots/small projects
IMHO, if you're that worried about resources then VM is not the way
But if you're learning how to setup VM then that's fine
no i'm not all at all tho but i don't want issues
Yeah, it will depend on how many people is accessing the service honestly
half of it is bots 😭
17K in 30 days that hits your server should still be within the capabilities of your allocated resources
so should i keep it?
it still to be honest better than my laptop when i hosted stuff on it
because my laptop has 2 cores and 1,smth i dont remember ghz
and that hp where i switched has 3ghz
Think of it this way. 17K request within 30 days is just shy of 600 request in a day. That means around 25 request/hour. So your if you wanted to check if it is enough or no, you can just test it by accessing the site yourself. If you feel any lag or it is not to your expectation, you can check where the bottleneck is
no lag
2secs to load
(its light php page tho)
sometimes lower
Yeah, again, if you're satisfied with it then it is good enough. You can test it using some benchmarking tools
Keep in mind that I assume the load is spread equally across time for that 17K
That's where the 25 req/hour comes in
install anubis and make those bots pay (in compute)
whats that
a proof of work captcha
Aren't CF providing pay to crawl now?
makes the client run expensive javascript to proof they are legit basically. but it mostly just slows down bots a ton so they'll probably just stop
i think its in beta still and you need to sign up for it
i use cloudflare tho doesnt it have that already?

i use cloudflared for sites bcs my isp doenst like portforwarding on 443
cloudflare doesn't do anything like that yet.
cf has their maze at least, even better
Make them think they're visiting an actual website 
can i use anubis and cloudflare?

alright i will try it out thanks
anubis will tho also screw up your indexing so it might not show up on google or something
since... they also are affected by it. not sure if you can configure it to not do it
is there way to fix it tho? i want allow only google duckduckgo and bing aswell discord for embed
Good ol honeypot
'ol reliable
idk. never used it.
"Anubis is a bit of a nuclear response. This will result in your website being blocked from smaller scrapers and may inhibit "good bots" like the Internet Archive. You can configure bot policy definitions to explicitly allowlist them and we are working on a curated set of "known good" bots to allow for a compromise between discoverability and uptime."
it seems like its possible to configure tho
I think anubis does have some sort of allowlist
impacts perfomance of server too? tho?
just asking
then wdym until i look at my hosting bill
because its gonna be making your hosting bill cheaper
well.. maybe 
it basically reduces traffic so it will also reduce your demand on the hardware
can i run anubis on other container or vm so if my first vm is down its gonna stil work?
read the docs
like i said i haven't deployed it myself so i have no clue about the details...
but probably
it just sends the client a payload to compute and then you get a pass or fail back
so.. no reason it shouldn't be possible to run seperately
that's an unfortunate typo
Based tbh
The theory is you're billed by request to your server. If your server experiencing a load of bot, then installing anubis will cut them down, hence lowering the bill
Since hosting (storage and idle time) bill is usually so cheap that they just bill by request
i'm gonna get anubis then
idk if my isp is charging me for hosting bill lol because my server is in my house
I think it's not a typo. Breaking changes because then the bot will have to adapt? But if it is about config, yeah unfortunate
Yeah, no they don't charge you. In fact if anything, ISP charges for bandwidth
anubis will lower my bandwidth?
idk if thats bad ngl
533.66 MB is what is taking of my internet?
or what
Yes, around 500Mb goes through the origin (your server). But for residential ISP they don't charge by bandwidth since the usage is usually within reasonable limit. You watching a lot of video is more straining than those 30 days of server running
i watch alot of movies
so prob disney took more than my server
and netflix
Indeed
i do think its still gonna help with performance
should i get docker or native anubis
i do think the docker version is the indented way of using it
Yeah, because then the request will not go through the relatively expensive computing service, but will go through some other check first
https://anubis.techaro.lol/docs/admin/installation/
everything should be explained in the docs
anubis should be in the front of any routing
because "COOKIE_DOMAIN=site1.com" i have like 4 site
so i do need per site set it or idk
i mean.. im sure its possible
but even if not.. running like 4 anubis docker instances is probably not that difficult to setup
docker compose will make that easier
One instance of Anubis must be used per service you are protecting
From the Anubis docs
subdomains counts aswell?
no
thanks god
Though
@rare bridge
There's nothing stopping you from having Anubis in front of another reverse proxy I guess
This is mostly architectural decision
https://anubis.techaro.lol/docs/admin/configuration/subrequest-auth/ I think you want this, so it sits between nginx and the actual apps
Anubis can act in one of two modes:
But I think placing anubis BEFORE the proxy is a good choice for this use case (low workload server)
Yeah that works too
how do i do that
Uhhh, by pointing the entry point to your server from cloudflare to anubis. Then the anubis will pass the request further to the reverse proxy (OpenLiteSpeed)
alright i set that like that
version: "3.8"
services:
anubis:
image: ghcr.io/techarohq/anubis:latest
container_name: anubis1
ports:
- "8923:8923"
environment:
TARGET: "http://(localiptoanothervm):443"
BIND: ":8923"
BIND_NETWORK: "tcp"
COOKIE_DOMAIN: "site.tld"
COOKIE_SECURE: "true"
DIFFICULTY: "4"
SERVE_ROBOTS_TXT: "true"
REDIRECT_DOMAINS: "site.tld"
but i wonder if isp wont see traffic to the local ip with stuff
anything behind a reverse proxy (or Anubis, same thing) should have an internal IP address in your internal network
your ISP will never see any internal traffic in your network, it never goes through any of their infrastructure
in theory they could send traffic with a local IP in your network as the destination, but in practice they never will and any properly configured firewall/router should block that
should i get ssl self signed for the other vm
so if anubis is proxing the :443 on another vm nobody could sniff what it does in requests
you could, but there's no point really
just make sure your internal network is secure 
the usual configuration is to let your reverse proxy do TLS termination for your internal services which use HTTP
chat
why
what if someone uploads smth harmful on my site and its gonna be seen in requests
Until zero trust comes into play 
idk i just made ssl to my server to calm down my paranoia abt that
paranoia about what
For some effective security, you need to know your threat model. What are you actually afraid of happening?
the most abt people hacking or uploading not legal stuff
Why would you have a open upload on the internet for no reason?
i don't open upload
but knowing people and bots people just can spam post requests with these
Hacking comes in a lot of variety. There is something called man-in-the-middle (basically like wiretap), there is phising, there is SQL Injection, XSS, and a lot more. Most of it is usually not of the concern for the end user but rather a developer if you're developing a program
For the most basic things that can be considered "secure", using TLS and having a good authentication scheme is usually enough
sql injection and xss should not be a thing at all in 2025 tbh
if the requests are internal to your network they will never be seen by anyone except you already
if someone else has access to your network you generally have bigger problems than whether you users requests are encrypted or not
but does there reqs save somewhere?
if you log them
like if someone sends smth harmful like 5 hours ago for example it can be visible still? to wireshark and etc
i do not
no, the reverse proxy may save request metadata, but only if you tell it to (and HTTPS won't help there)
so i dont have to worry that if someone sends to my server requests with idk gore or smth it wont save on my server? if i dont log that
thats not how that works
wireshark is phone tapping basically, if no one is touching the wire nothing gets intercepted
idk abt cybersecurity stuff too much 😭
like type of that
you only have to worry if you serve illegal/abusive content to users
even storing it shouldn't be an issue (unless it's personal information or something, then mabe)
i do not serve these stuff if someone uploads even stuff like that to my instances i will just remove it after i notice it
should be fine then
so i shouldnt worry abt that?
no
...should
Tbf, I agree for SQL. Mostly talking about XSS stuff
should i use tls for mariadb tho?
you only have to worry if you serve content that is illegal or content that violates policies of whatever service providers you may use, like cloudflare or the hosting provider or the domain registrar. if content uploaded by users gets served immediately (especially if it's discoverable by anyone) you should be careful
actually, one exception
maybe don't reverse proxy those requests unencrypted over the Internet
but that's fixable by using a VPN like Tailscale, still no HTTPS required
if it gets uploadd i remove that type of stuff fast
tho it never did get uploaded yet
you’re generally good then
should anubis protect me from it too?
because i know there are bots posting these stuff on vichan instances
if your concern is bots then it should
my concern is abt that spam only https://github.com/vichan-devel/vichan/pull/780
(i mean problem not the unshorten url thingy just saying)
but you are not hosting an image hoster?
not at all but i planned to
to make small imageboard
well, have fun with that nightmare
also there’s this https://developers.cloudflare.com/cache/reference/csam-scanning/
reduce maybe, prevent no
does it block upload
or viewing
Does the image board require account to post image?
Because if so, you can just disable registration
no, it just notifies you when it finds known bad content
And only register via invite link
in my own coded software yes
Oh boy, have fun
it does actually
if your website is proxied through cloudflare, and it caches an image that later happens to match a database of such material they stop serving that image and send you an email, but you have to remove it in reasonable time or you get banned
if remove that do i need do smth other or no
does it stop the requests or block them tho
no it retroactively scans already cached content
I wonder if there will be chaos if I bring up the question "Which Linux distro is the best" 
microsoft windows
Debian
i’m afraid this but unironically
wdym
everything on my server runs on debian
like it’s objectively true that debian is the best distro and i’m afraid it is
forget linux, run templeos on everything

i don't trust that fish
it’s too big
the hell is this
roblox

I will not elaborate (this time
)
I know it is roblox duh, I meant the context of this site
roblox
...
roblox
you’re wrong because the best linux distros in order are: microsoft windows, mac os 7, android

best distro is Ubuntu 
that last one might actually be real
it is technically one
Hannah Montana linux ftw
Arch with Snapd
TempleOS better
yeah, and it's the only one with users 
Android is becoming more locked down, so it is getting it's linux rights removed
i'm gonna make better nvidia drivers from scratch in holyc
ok
are you lord 
Android has become to Linux what macOS is to FreeBSD 
morning
woke up from the loudest thunder i've heard in years
The worst part is, if you root and install Linage, you are locked out from a lot of apps
for no fucking reason
fucking McDonalds app has root detection 
I love sleeping in Thunder 
it was reverberating for like 20-30 seconds
wait until 📸 📸📸 🎆 🔊 🗣️ 🎆
You sure that isn't some heavy machinery going down?
Or some Super Weapons being fired I guess 
nope its thunder
i live just a tiny bit too far from the frontlines to hear any weapons
I sure hope not
abwabwabwa
Awawa?


is this an ARG

playing /dev/urandom in ffplay: is this a neuro arg reference
urandom has pattern
I choose /dev/random
they're literally the same thing
no
only if you stream too much
urandom gives as much data as requested and random only gives as much as is currently available, so may error if there isnt enough in the entropy pool

mhm so reading from /dev/random can block until there are enough random bits of entropy
urandom is identical to random provided that there are enough random bits of entropy
the difference is in the behaviour if there arent enough random bits
wait this may not be true in modern kernels though 
It is not true
yeah looks like in modern kernel, they are functionally identical
guys i wanna selfhost my own git should i use gitlab,gitea,forgejo to work well out of box
with urandom being preferred
not anymore i think
After the seed is initialized yes
between gitea and forgejo i would always choose forgejo
gitlab is kinda big but works great
otherwise forgejo
gitlab has different features though
hm maybe not https://lwn.net/Articles/889452/
In mid-February, we reported on the plan to unite the two kernel devices that provide random nu [...]
/dev/random only blocks when it hasn't been initialized
That change made it as far as the mainline during the Linux 5.18 merge window, but it was quickly reverted when problems were found
typically forgejo is much more lightweight than gitlab. gitlab has "more" features, but they may not be needed if you only care about the git part
Oh, its Urandom not Random that blocks
what are diffrences between forgejo and gitea
forgejo is a fork of gitea
no you were correct /dev/random blocks
in the open source space, probably because forgejo has a more open license than gitea
Oh it is. Dang, multi-tasking is hard
after this forgejo was created
its also maintained by more trustworthy people (codeberg)
one is still more random than random
urandom takes in random as seeds

why is #programming talking about fucking version control platforms 😭
????
is this on bingo
the quality of the random value is the same
also any open source project that has this doesnt deserve to exist
meh they can be cracked by quantum computers under 10 minutes anyways
where is the irc
doesn't matter
do you have a quantum computer

All computer is quantum computer 
Its just manipulating/using the quantum state differently
yay 156 qbits
Goodmorning
to me this looks equivalent to "you can contact us on facebook!" wow cool i won't do it though
and i know this is ironic to say on discord but it makes sense to use discord for non-technical communities like this one
guys i think i fucked up anubis configuration
it keeps refreshing page or redirecting me into same page in loop
fix it 
did i set smth wrong
let me check what nix did to my config
nvm it was only cache of my browser
using librewolf is pain
cache exists in all browsers, but isn't it supposed to use etags
anubis mentioned?
yea but librewolf breakes half of sites
like which half
okay to be fair i have privacy.resistFingerprinting off
mf work
same, remember it breaking a bunch of sites when i had it on
i also do
somehow that thingy does that
this is a separate setting
Isn't resisting fingerprinting only makes it easier to track/differentiate?
the point is to throw all resistFingerprinting users in the same cohort
removing hardware details and leaving something generic
Yes, that is the idea. But it depends on how many are actually using it too no? Because the requester IP will still be visible anyway
of course ip is a separate tracking vector yes
Like how tor is only anonymous because if only the government is using it, then when a traffic is detected as tor traffic, it is known that a government is using it
how do i make that Accessible 😭
I honestly have no idea what dashboard is that
bwaa
OpenLiteSpeed
host forgejo
ports even wont work
Alright. Is the forgejo confirmed working?
yea
localhost:3000 works
no matter what port its always disabled for some reason
Have you configured forgejo to accept request from the reverse proxy (in this case openlitespeed)?
i dont think so
how do i do that
it blocks all ports btw
by default its always green (other vm)
I'm on mobile, so can't help much beyond asking some question. But for forgejo to works behind a reverse proxy, iirc it needs to be configured to accept the connection from the proxy. And then configuring the proxy to accept the connection from anubis and passing it along to forgejo
.
in theory it only needs to listen on the correct IP address, but you can easily test that by just using curl from wherever the reverse proxy is hosted
any configuration of the public IP/hostname is usually only for links and webhooks and stuff
requests should still go through without that, even if redirects might be messed up
Oh, good to know that the setup is a bit less painful than gitlab or other forge server that I forgot the name before
So maybe the VM firewall?
the only forge that has been a nightmare to try to set up so far has been sourcehut
I failed
should be painless on nixos
also do you know about https://gitlab.com/simple-nixos-mailserver/nixos-mailserver/
the listener isnt just running for some reason
have heard about it 
I'm not planning on hosting my own mailserver though


the last time I tried was already on NixOS
the package still existed but was already not in a great state
there were issues™
Why are you using 9999 if the forgejo is listening on 3000?

Should I sleep or embrace more caffeine in my system?
embrace caffeine and sleep
Ebough bedrotting for me, i need to get back to work 
mrp
Sleep more. You never know when you will not have that privilege anymore 
I 1lready slept 10 hours
I hate how i charge my slartwatch to track my sleep, i go to bed, and when i wake up it has 2%
Like bro wtf
And it didnt even track my sleep cuz i forgot to unlock it
i don't think that's supposed to happen 
Well it did
Man smart watches suck
I only hear them being good as phone extensions and otherwise dodgy at best.
I get iy only has 247mah, but man the display is only on for a couple minutes per day
being able to have my phone on silent is kinda nice too i guess
Use a proper smart watch if you want. Like Apple or something that runs full Android
What is the android name again? Watch OS?
Man, tech companies and their naming sense
smart watches are convenient but i'd never buy one for the full price they ask tbh
I have a galaxy watch 4. I would consider that a "proper" smart watch?
i have the same one, tf are you doing with yours that it's not even lasting a night
granted i have the classic but still
Idk
I have the same model
the galaxy watch 4 is kinda weird since it's like half wear os half tizen iirc
How long have you had it?
2 years
Yeah, but better than some chinese cheap "smart" watch one I think
Sure
only 1%
Huh, the battery is usually still good for that time. But maybe your battery is more degraded somehow
it does watch things 
But not so smart now are they 
I havent updated it in a long time tho
Lemme see
Oh i need to update the app first, goddamnit
i'm still annoyed about that one time they pushed out a major update and it changed to require an extra tap to turn on sleep mode since there are more modes now
never have i ever used any of the other available modes
wtf am i going to do with a "theatre mode"
I have never used sleep mode lol
I used it twice
I had to go to theatre for dchool
And i went to watch a movie
i just turn on sleep mode when i'm in a theatre
When you touch grass 
all it does is turn on dnd anyway
I think it also blocks alarms?
So you become a dungeon master or the player? 
yes
Nice answer
Guess il see if its better
A throbber, also known as a loading icon, is an animated graphical control element used to show that a computer program is performing an action in the background (such as downloading content, conducting intensive calculations or communicating with an external device). In contrast to a progress bar, a throbber does not indicate how much of the ac...
hey, can anyone give me a direction here? would focusing on javascript be a bad idea if i were to lean towards AI related specializations?
python would probably be the better bet in that regard
Unless you're talking about the interface part
Nope, absolutely not
Javascript is for web, if you want to do AI related things learn Python and C++
well i haven't looked into it much since i'm focusing on uni topics atm, but i was interested in AI and my sister told me i should learn how it works so i don't fall into the trap of learning how to train it. She said i could learn either python or javascript, with python being safer but javascript being doable as well.
personally i prefer programming anything in javascript, but idm python
figured i'd ask an AI related discord about it
Javascript kinda sucks
if something.. typescript would be better than javascript 
The trap of learning how to train it?
So, do you want to do the AI stuff that is the backend things or the front end things?
basically:
- running ai = python
- training ai = python
- making new ai = c++ (or any other low level language like rust etc.)
no, making new ai = python too
has been a while since we had that convo, but from what i remember is just so i don't become the guy training the agent, but the guy making the ai
you don't write new ai in python 
you do
you use libraries like torch that expose a python dsl that compiles to gpu code which runs way faster than something written in rust would run on cpu
i'm more inclined to backend
why would rust mean it runs on the cpu 
but i'm also still an uni student so i really don't know much about what that would be
it does not, my point is python being slow doesn't matter for AI
well technically if you really want to go indepth. you would need to write cuda compatible code. and that is likely just easier in languages like c++ or rust or something. you won't write a attention head in python. that just sounds.. weird
interacting with ai and putting them together into something useable is done through python im sure. but you won't have the most access you will need to actually be able to write proper new ai's
Uuh, please clarify when you say making the AI. As in, designing the model internal layer architecture?
i assure you pytorch is all you need
the field is also ever changing. whatever we say now might be obsolete in 2 years
this only applies for truly novel architectures and/or optimisers, for the "common" ones there are already efficient implementations readily available to call from python or any other language really
i doubt something like deepseek or something is made in python
im sure its a big part of the pipeline. i don't doubt that. but lower level languages are still insanely important. (assuming we are talking about NEW AI. not variation of existing ones)
usually you prototype in python and then the final deployment gets rewritten in c++ or sth to then run
its not made in python not because you cant make it in python, but because after the prototype stage you have to optimize
This needs to be clarified because for you to know the performance of the model you just designed, you need to train it and evaluate its performance
python is a turing complete language... you can make anything in python
i don't really understand what that means fully, do u mean the code which encompass the process of understanding of the AI?
let me repeat again, the code that's being ran isn't python but a python DSL that compiles to GPU code
and... i get that. c++ would still be better tho 
not necessarily

I mean what do you mean when you said making the AI but not the one that trains it? Because as I said, the terms AI is too loaded with ambiguity from the people outside of the academia
i thought those were two actions done by different devs, the one designing it and the one training it. i guess it could be done by one
the way i thought it worked is that team A designs it and team B trains it
always those damn mathematicians 
Uhh not really, no. For large AI system that needs distributed computing it may be that way
But even then, it should be the infrastructure team that provides the rnd team the necessary compute resource to design and train the "AI"
If you mean don't become the infra team, then yeah, learn python
Evil is live yeay! 
hm, i'll look into it more and talk to her again so i can ask clearer questions, but for now i'll take the advice on python, ty 🙂
clearly the solution is to write mlir by hand 
python never hurts to know if targeting the "AI" sphere
you will have to deal with it one way or another eventually
great, at least learning it won't be for nothing i hope
every second i spent learning java i felt like dying inside
as much as i would honestly love to say learning python is a waste, it's not 
python is a lot more lax than java, for better or worse
java, i didn't
After you learn a few different languages, they're all the same
if you're not actually learning different languages, they are
talking about python makes me wonder if i could possibly persuade my thesis advisor to let me write my bigint library in mojo, that could be fun 
though ig it's still a very new language so eh
I'm just using a programming language as a tool like any other tbh 
Don't wanna write raw machine code all the time
i love a good abstraction 
i may not have a brain but i have an idea

:hmm:
hii
hiiii
why forgejo feels so much faster than gitea
the proxy was lsitening that http port not forgejo
probably is
thought my reverse proxy is pain because sometimes it shows 403 for no reason
or bugs the anubis
it's a sign from above to not use mongodb 
i dont use mongodb
poor anubis girl

i'm switching mongodb to PostgreSQL in some projects

it doesn't matter, choose whichever one you prefer
i mainly mained mariadb, but uh idk if its anygood people tell me its outdated
that being said, since you seemed to care about resource utilisation earlier
postgres is considerably heavier than mariadb iirc
and slower
it's fine
and i need fast database for one project the fastest possible so uh
mongodb kinda sucks in my experience
it doesn't matter unless you're google or sth
i'm making spamdb discord bot what checks if user is in database then bans them
i stand by what i said 
but for that i think db needs be faster
besides, if you need the lowest possible latency you want something more akin to valkey anyway
idk if mariadb is fast for that type of stuff
nah you'll be limited by the discord api
not sure why you would want to but there you go
idk speed
and being light?
nodejs was kinda pain also idk if its any good for that type of stuff
being light is fair ig, as for speed it really doesn't make a difference unless you plan on supporting commands that do a lot of number crunching 
i want it to just use less resources
then have fun 
you do give up a lot of convenience, whether that's worth it is up to you
it seems there's no certain type of infra/backend nerd here... shame, i wanted to see hammers fly into your general direction because it scales badly or something
if postgres can't handle your needs then you're either incredibly lucky to be in your position or you're an idiot who can't use a database tbh 
imo for discord bots development velocity matters the most, when i had one my bot supported hot reloading
(in a very bad way where it just read the command from the disk on every execution but the scale allowed that)
also speaking of postgres, TIL planetscale supports postgres too now interesting
what
i prefer postgresql simply because everything supports it
(almost everything)
sqlite for small scale works too, and at large scale sometimes nosql works best
just compromise and use mssql
postgres takes up considerably more resources, since you care so much about that
taking up resources isnt bad if they are used for good reason
if
mssql?
Microsoft SQL Server is a proprietary relational database management system developed by Microsoft using Structured Query Language (SQL, often pronounced "sequel"). As a database server, it is a software product with the primary function of storing and retrieving data as requested by other software applications—which may run either on the sam...
how tf i'm supposed to use mssql on debian
they updated the logo huh
ok turns out they are making their own vitess but for postgres, crazy - and they don't "support" it, they just run it on their infra without scaling glue yet
also very limited access
how perfomance is in it
no idea
i will just use mariadb
if it ain't broke don't fix it 
vitess for postgres does sound pretty nice huh
my bot needs recode tho
codebase sucks asf
accidentally cmd-q while focused on discord but internet is dogshit and discord downloads the whole app bundle on macos when updating so like 200-300mb, gg
and then taken over by planetscale a couple years ago
is it
ye you’re right but i wasn’t sure on this
makes sense
mrrp
konii is cat
ngl should i keep bot coded in nodejs, switch to python or rust
are you comfortable with node? then 0 reason to switch
discord.js is by far the most well-maintained option of the 3 so there's that too
i meant my src is messy
it needs recode
at some point
then what you need is a refactor or a rewrite like you suggest, not changing languages
why does it need a rewrite if it works
"works" it crashes alot sometimes
also its hard to me to update it due to code structure
also dashboard is fucking mess
and i dont feel comfortable using the uhh mongodb because i wanted try it out but uhh i didnt find it good for me
true, meant to sat "what you want" not need
also i hardcoded ADMIN_USER_IDS to code
and every time when iwant add someone or remove i need restart bot
moving my screeps codebase to use more generic types like storable & buildable and having generic fuctions operate on them instead of me re doing it all every time, simplifies the code just a little bit...
Sam not even needing konii to trigger it, smh

i like using pm2 personally but that's just me
if you're already using docker i don't think it matters too much
is there better way to edit src's inside filezilla sftp
i find it harder to just edit some part of src
Most code editors have a way to let you edit remote files over ssh
vscode can just connect to a folder on a server and directly edit stuff there
vscodium doesnt have that?
probably not. because it also needs to install a service on the server i think
pretty sure the ssh extension is proprietary and wont work with vscodium
There's an open source version of it but idk how well it actually works
I guess there's also nothing stopping you from just adding the official plugin marketplace to vscodium 
thats not the issue
even if you add it it wont work
because it relies on proprietary components in vscode itself
well at least thats how i remember it
Idk
I'd test it rn but I cba to setup vscodium
Probably should though
Either way, there is a replacement extension you can just use 
Most MS extensions are limited to run on only MS products by their license and by running additional checks in their proprietary code.
sad
why tf is latex workshop on that list
i love emacs tramp mode for this
apparently if you have a US international layout you straight up cannot type "`~ in zed 
except i can't reasonably do collaborative editing
Imagine inversing or using markdown
it was thankfully fixed in a pr 3 days ago 
New nickname just dropped
why am i bullied here too
Sometimes 😭
poor vanor




i would respond but automod would probably put a shotgun to my head
rephrase and send again 
isnt that also against the ruiles?
hmm i think theres a rule for this 
rule 12 is not real
lets just say i support workers and move on
it cant hurt you
Something about no loopholes 
rephrase is rephrase 
surely vanor isnt still watching 
a loop is topologically the same as a straw and therefore it has two holes
wrong
Nah, Vanor is probably getting busy being bullied in NN
a star and a loop both only have 1 hole
everything is donuts anyways
straws dont have two holes
straws have one hole
topology
schizo
in topology, everything can be mapped/transformed onto/into a donut, thus everything technically have only one hole
problem solved
Sam 5am, my brain doesn't see the difference clearly 
what about trousers
they have two holes
it's consisted of multiple donuts 
i watched a vsauce video about this but the video starts with him on the toilet so im not sharing it
Wow there is indeed a drawing for this
the human body is 7 holes
theres more than just the number of holes that is inspected in topology
you beg to differ? then beg
-# There are also warcrimes in topology 
the sinuses are complicated
airthight = blind hole
the mouth and nostrils are connected and therefore i posit that there are only two holes on the face

if you think about it nostrils and mouth are like trousers
mouth is a hole, each nostril is a hole, and you got some tubes to your eyes
but they are all connected in the sinus cavities
Connected to mouth 
yet can be morphed to seperate holes
you got 1 hole from the top to the bottom of the body
2 nostrile, 2 to the eyes, and 2 more for the tears
i dont see it. loosely the nose, eye ducts, mouth, and anus are all connected holes due to sinuses
-# Hi mods 
you're not thinking as a topologist
also the eyes can be morphed to flat
i am. its the same situation as trousers. 
and eyes(lacrimal puncta)
the eyes can, the ducts cant
duh
but the ducts are connected to the sinuses. you subtract one total hole because of this.
What a great break from writing not yuri
topoligically correct smoker
read the about section
very technical conversation I could tell
this is very technical and important for human simulation research 
googling "topology human" i should have expected the results 
what did you get
blender woman
ahhh
aaaaaaaaaa
we love topology
me after writing without any mental breaks all day
a hole connecting to another hole instead of "the outside" is still a hole
so the main sinus cavity thingy will "belong" to one of the holes with the other one ending there
(I still have 4 hours left to write
)
I blame myself for stalling so hard
https://fxtwitter.com/hanagamigendai/status/1945996943735627794 computer work be like
i have worked 3 days in my life so i wouldnt know
I'm hopefully going to be first year soon 
what gamejam 
shovel game jam
nope

me, iggly, superbox and weegee
6 of the 9 days have passed, and i barely have a functioning demo


will yall be joining the gmtk as well?
3 days of fixing the engine, 3 days of 3d modelling something that was probably too big of scope so i canceled half the map
it's on july the 30th
Nah I'm perishing after thiz
In actuality, I'd just want to work on the parts I skimped on.
behold my opengl
More yuri
id rather not, opengl is sucky and cant even push sub 0.1ms frametimes
think about trousers though. it appears to have 3 holes but topologically it has two because the third hole doesnt actually exist
the fake end is the part that connects to your mouth
that just means you don't have enough cpu clock speed
nostrils, mouth, lacrimal ducts are all connected hy the sinuses, which is then connected to the anus. then you can model human as a sphere with 6 openings which is identical to a circle with 5 openings
im pushing 40% gpu usage and 4% cpu usage.
its mainly opengl's overhead using 70% of the frametimes
bandwidth limit 
like actually when i open the game on vsync my pc doesnt even notice
nope, just opengl being bad
do low-level controls over opengl
thats called vulkan
sounds like a waste of time
you're describing vulkan
vulkan is more portable than gl to begin with
That's called Zinc (vulkan again)
that's minecraft for you 
port minecraft to vulkan

immeditalt quadruple fps
As the resident minecraft nerd, no
16 year old game has tech debt 
theres a mod for this
why?
forget opengl and vulkan, I choose DX11 
I've tried this mod, conflicts with every performance mod known to man.
i cans ee why
only ones that touch rendering
i cans ee why
the performance mods proabbly dont expect a completely diffrent backend
Sodium:
things like sodium optimise minecrafts render pipeline
a pipeline that doesnt exist when using vk

not the same pipeline
it's not replaced, it's patched over 
Sidenote, I did not know neoforge pulled it's act together and actually now has the promised fabric support
vulkan uses the same pipeline setup as opengl, you just have to make each one manually while in opengl you only make the vertex and fragment shader manually
for minecraft tho you'd be better of with a mesh shader so that is the only way you're using a completely diffrent pipeline
I was surprised to see a neoforge mod with a fabric mod dependency
minecraft does most of its rendering in immediate mode 
if anyone has made a voxel engine before you know that this is a performance killer
immediate mode just doesnt use a vbo but its actually bad for performance
hold my VBOs and VAOs
not that type of immediate mode

awa
awa
I'm extending my break to run a new minecraft modded client
qwq...?
by run do you mean downloading and executing, or developing a new modded client
sure
i would say just make sure you're finished in time but that would be hipocrytical
`(*>﹏<*)′
I mean I stalled for 6 days, I deserve it
idk if im gonan be able to finish this in 3 days
don't ignore me 
I did not ask for you
I usually assemble modpacks/modded clients by hand
I see
curseforge is where I find my virus
Seems like the boi is not vibing
probably forgot some vulkan package on my system
sidequest for later did someone say RT Cores https://modrinth.com/mod/vulkanite-mod

RT cores?
where?
idk about vulkan packages, i thought that was jsut part of your gpu drivers
I forgot a package 👍

kinda need vulkan-tools to use vulkan
nope, MC is stroking without me interfering
ohhhhhhhhhhhhhhhhhhhh the fabric loader is bugged
I had to knock out half of the mods in the client but this is with vulkan-mod (menu would not shut while taking screenshot)
is the performance decent?
im assuming 144hz is vsync so not using full gpu?
oh rip screenshot
redoing test, this client did not have args
ah
(pretend this is 140fps)
uncapped fps, saw a 200 for a second this time
120 while moving, so I'm considering that the min
Now time to swap back (note, no nvidium due to iGPU)
also not distant horizons becase that preforms worse on this system
lower fps while moving makes sense. i had a time where moving made the fps drop to a 4th
shaders time :xdx:
slowly but surely solving the housing crisis
a test I couldn't do on vulkan due to no support for shaders
i have discovered that usign the same model twice breaks collision tho
more debugging
Complmentry, default high settings
I do not want to migrate to windows 11
and take this sunset
then dont
its playable-ish
i want the new updates tho
you just said you didnt
secutirty updates and feature updates are two very diffrent beasts
ye.. its pretty much a package deal so 
such
moment ngl 
what im about to do is probably really stupid, and wastes vram, but i dont have the time to make a better system for this
if using the same model multiple times causes collision issues, ill just make them seperate models of the same file
-# This pack does not need 16GB ram
how does this happen
do instances not get separated or something
like this
rendObj.model gets changed per instance
cuz each instance has its own boundingbox
or its supposed to have that
oh so you give each instances hitboxes
actually this shouldnt take too long to fix in the proper way, ill jsut move the boundingbox from modle to object
i get the hitbox data straight from the gltf file, but i still need to put it from object space into world space
Alright 2gb is good enough
2gb 20 render distance is crazy
forge moment
fabric actually but yeah
you're on diffrent pasges



