[ overboard / sfw / alt / cytube] [ leftypol / b / WRK / hobby / tech / edu / ga / ent / 777 / posad / i / a / R9K / dead ] [ meta ]

/tech/ - Technology

"Technology reveals the active relation of man to nature"
Password (For file deletion.)

Matrix   IRC Chat   Mumble   Telegram   Discord

File: 1626055997034.png ( 291.47 KB , 485x436 , 1623078444647.png )


Hey /tech/ I'm here to say I'm probably going to be rewriting a chan in a new language/stack intended as a vichan replacement. I've reached out on lainchan:
and I'm also reaching out here. Initially I was writing it in java, but after getting some feedback from other people I've decided to take a step back and solicit more feedback from the userbase and people here on /tech/ and in the wider alt chan community on what it should be.

What does /tech think of:

Architecture: Monolith vs Microservices
Front end: SPA (ex: Angular, React) vs Server Side HTML templating
Backend Language: Java, C#, Lisp, Rust, Golang, PHP, etc.
Database: SQL vs NoSQL


As long as you don't force feed users webp, you have my support.


What do you have against WebP? The only thing I don't like is how some CDNs and webmasters name their files with ".jpg" file extensions but serve WebP anyway, that is deceitful and just plain annoying.


Arch: Monolith
Front: SPA
Back: Whatever, except for Java and raw PHP

SPAs are so much easier to code non-basic features with, I don't think I can ever go back to server-side rendering


Your previous thread got me thinking about what I did not like about your java experiment btw I was obvious joking about it being enterprise language. if you are serious about your project you can totally make it deployable on cheap VPS by reducing/eliminating memory allocation after application startup. I know of few trading desks that writes java applets like that. It's bit too much of a work for side project but it is technically doable. It's not about a language but architectural direction you are taking. Would be nice if you elaborate on how your project will depart from vichan.

I liked that you were aggressive with server side rendering but for application that's sole purpose is being dynamic like imageboard, I do not understand what you actually gain this way. Keeping backend as minimal and succinct to the point of REST API point that handles user's posting seems more logical on surface level but I am not putting too much thinking into this right now so I must be missing something crucial. This design will also allow admins to make it much easier to federate one another and make frontend design choice completely decoupled which could be good thing and a bad thing. WASM is powerful enough to host complex image manipulation routine in browser side and I am confident there was even imagemagick compiled to target wasm at one point so things like thumbnail generation and other quality of life processes can potentially be handled by users

On the choice of DB, stephen had such a strong opinion on adhering to mongo for some reason. He claimed that he experimented with postgresql's supposedly enhanced json capabilities but were not impressed by it. Instead of theorycrafting one needs to do quantitative benchmark but that would be difficult without deciding on backend architecture.


My heart says lisp
My brain says you should look at BEAM VM languages like erlang or elixer


Does a Chan for shitposting really need telecommunications system level reliability??


> telecommunications system level reliability
Pleroma is written in Elixir and runs on OTP and it is broken all the time.


well whats your suggestion then?


Use whatever language you are most comfortable with. Wrestling with the language instead of the problem at hand is a sure way to trouble. If your goal is making a vichan replacement, focus on achieving that before you think about all the new revolutionary nonsense that you could be adding. Make it simple, most altchans are only visited by three people at most and I doubt any will ever get as popular as 4chan. Don't try to solve problems your users don't have.


File: 1626072882707.png ( 551.99 KB , 1062x649 , thumbsup.png )


>except for Java

Why not java? it's got libraries for everything. And its faster than most languages except low level ones like rust/c/c++/golang. It's also got a low learning curve which is why many universities teach it as a first language which as a consequence now tons of people know it, and if they didn't learn that they learned C++ which is very similar to Java enough so they can easily make a transition. So the supply of people who know Java is pretty high and its easy to learn too.

Uber, netflix, airbnb, pinterest, instagram, spotify, amazon, slack, obviously oracle and even google search was originally written in java. Banks use Java, Universities, and fortune 500 companies.

Spring boot based services is a perfectly acceptable choice for any "startup". Its a jack of all trades language, I'm always puzzled when people choose the hype meme languages over a proven workhorse like Java, instead they say you have to do something crazy like write your site in an obscure dialect of lisp or erlang or something. You can't beat the rock solid stability, tooling and maturity of the JVM ecosystem and JVM itself.

Vichan derivatives have powered chans for more than 10-12 years now, imagine if the imageboard of the future gets written in a hipster language that is obscure in the future, imagine if vichan had been written in ruby on rails.

People's real complaint seems to be they want an "exciting" language to write shit in, but the older and more tired you appreciate an ecosystem/language that is mature and conservative and not constantly churning with activity and breaking compatibility creating more work every 6 months.

>Use whatever language you are most comfortable with. Wrestling with the language instead of the problem at hand is a sure way to trouble.
Yeah thats the conventional wisdom but the problem is that to get buy-in from potential devs on an open source project you need them to want to develop in the language you've chosen. So actually the choice of language does matter in that regard because if its too hard of a learning curve they won't contribute because they can't/don't want to.


highly based


People who want to contribute will contribute regardless of the language. Most people won't contribute and it is pointless to try to please them.

If you want people to contribute it is more important to make your project easy to set up for development. Especially for an imageboard where you will need a webserver, database, some test data, etc.


If it helps you make a decision I've been contemplating writing an anonymous textboard in erlang for a while


File: 1626074941893.png ( 326.69 KB , 633x692 , lains2.png )

I guess for this I will do a strawpoll to get peoples approximate feelings on this.

I'm doing a language survey, based on the tiobe top programming languages, of what you guys think would be good to write in. Obviously, excluding non relevant languages like C/C++ assembly, and languages that will likely be used anyway like SQL.


lets see what people think first off.


Java just doesn't feel fun though. And with Spring, even if you know Java, you have to know Spring very well to do stuff.
With something like nodejs+express almost all you have to know about express is how to register a new route (1 line of code). It's easier and provides more fun. Plus very good environment.



>And with Spring, even if you know Java, you have to know Spring very well to do stuff.
>With something like nodejs+express almost all you have to know about express is how to register a new route (1 line of code).
but thats because spring does stuff for you automatically? in any case its pretty easy to learn in any case


looks like java is winning the strawpoll


For modern webshit with loads of js activity, I always recommend the NodeJS/NextJS combo.
Does pretty much everything you want from a modern web UI and easy to use.
I don't like it for personal stuff because I'm an oldfag who likes my simple static web 1.0 stuff with no surprises, but anyone starting a new web project that needs to have mass appeal, this is the way to go, would allow you to do all the shut zoomers expect from a website with minimal headache.


Kill yourself go make a video game in python with your indian college faggot dorm fuckbuddies


>le choice of language
You will never get anywhere, just give up


idk why you're necrobumping but yes coding MUST be fun for side projects otherwise you will get nowhere.


What you are saying is correct.

However a surprising amount of code gets copy-pasted from stack-overflow and then tweaked to fit the application. Recently people have begun a similar praxis by prompting large language models, and tweaking the output of that. Neither is particularly fun.


ChatGPT has made programming way more fun for me. I'm blocked a lot less often, I can delegate the boring stuff to it and focus on solving slightly higher level problems. It's been a game changer tbqh.


>ChatGPT has made programming way more fun for me.
I haven't tried it to be honest. Tho I watched online videos of people using llm code. It didn't look compelling to me. People spend as much time refactoring the generated code as it would have taken them to write it from scratch. To be fair it's been a few months since i last looked at this, it might have improved since.

I could see this as a great way to learn a new programming language, and possibly as a tool to detect bugs and perhaps security flaws.

I'm not sure about using a online llm services for production software tho. I would be too paranoid about the code-generators being used to slip malicious code into programs. Consider that Code-llms scrape all of github, which means that if somebody figures out how to spike "delicious" code-snippets, that the llms like to pass on to users, they only have to upload that to github and then loads of people get pwnd. If i wrote programs for other people i would use a airgapped computer, and then upload the finished software from a connection-point that is not predictable. For fears of getting hacked and abused as a vector for distributing malware.

Security conscious developers worry allot about supply chain attacks and go to great length like encrypting the ram on their machines because that somehow makes it harder to compromise software at the point of compiling binaries.

the FUNmetric
My take is that "llm-prompt-coding" will be fun once it becomes more like a sudo programming language with predictable output and less like a code lottery. My concept of fun is playing the same game over and over while incrementally improving my skill level until i reach a plateau. So the code lottery aspect looks frustrating to me.

Unique IPs: 16

[Return][Catalog][Top][Home][Post a Reply]
Delete Post [ ]
[ overboard / sfw / alt / cytube] [ leftypol / b / WRK / hobby / tech / edu / ga / ent / 777 / posad / i / a / R9K / dead ] [ meta ]