Conversation

Notices

Moon (moon@shitposter.club)'s status on Monday, 29-May-2023 23:31:17 JST Moon
@crunklord420 @alex @graf @lain @mint I agree btw that nobody (afaik) said to do this before now. Well, I did but I didn't take my own advice for mediaproxy, but only because I thought it would be harder than it was because I didn't see the option in the docs (maybe that is because I am stuipid though.)

In conversation Monday, 29-May-2023 23:31:17 JST from shitposter.club permalink
- matrix07012 :thotpatrol: :cunnyEmpire: likes this.
- | (n-2-l@rdrama.cc)'s status on Monday, 29-May-2023 23:44:07 JST |
  in reply to
  @alex @lain @graf @crunklord420 @mint @Moon every single javascript engine is slow as shit,nodejs is written and C and uses libuv underneath.it doesnt matter if the runtime is written in rust or whatever because the js that runs on top of it requires gigabytes of RAM to do anything because of so many layers of garbage thats tangled together.
  
  In conversation Monday, 29-May-2023 23:44:07 JST permalink
- Alex Gleason (alex@gleasonator.com)'s status on Monday, 29-May-2023 23:44:07 JST Alex Gleason
  in reply to
  @n-2-l @lain @graf @crunklord420 @mint @Moon I'm bridging two entire decentralized social media protocols on a single thread of TypeScript code in Deno, serving about 4GB of data per day, and the whole VM including my code and OS are consuming about 400MB of RAM and maybe 10% of the CPU.
  
  In conversation Monday, 29-May-2023 23:44:07 JST permalink
- CrunkLord420 (crunklord420@rdrama.cc)'s status on Monday, 29-May-2023 23:44:08 JST CrunkLord420
  in reply to
  @Moon @alex @lain @graf @mint I can't believe I'm saying this.
  
  Pleroma needs to be rewritten in Rust.
  
  In conversation Monday, 29-May-2023 23:44:08 JST permalink
- Alex Gleason (alex@gleasonator.com)'s status on Monday, 29-May-2023 23:44:08 JST Alex Gleason
  in reply to
  @crunklord420 @Moon @lain @graf @mint I'm building a backend in Deno which is written in Rust and uses tokio under the hood.
  
  In conversation Monday, 29-May-2023 23:44:08 JST permalink
- CrunkLord420 (crunklord420@rdrama.cc)'s status on Monday, 29-May-2023 23:44:51 JST CrunkLord420
  in reply to
  @alex @lain @graf @n-2-l @mint @Moon almost everything I run is never CPU bottlenecked, and that's a bad thing. They're always IO bottlenecked, pleroma, matrix, everything. The CPU is literally just sleeping while waiting for data to move from disk to ram to CPU. If the software was good that data would be in CPU cache, it would be in memory, it wouldn't be on disk. It wouldn't be constantly doing system calls to the kernel and sleeping on mutexes.
  
  If the software was good, it'd be using 100% on all cores.
  
  In conversation Monday, 29-May-2023 23:44:51 JST permalink
- pistolero :thispersondoesnotexist: (p@freespeechextremist.com)'s status on Wednesday, 31-May-2023 18:19:25 JST pistolero :thispersondoesnotexist:
  in reply to
  @crunklord420 @Moon @graf @lain @mint
  
  > fundamentally speaking it will always be inferior to a non-VM, non-GC'd language.
  
  Bullet-point meme objections. GC's fine, Pleroma's not memory-bound. VM's fine, it's not CPU-bound either. It's I/O-bound, leans hard on Postgres. Maybe we should add manual memory management to Postgres, start shipping binaries to the DB instead of letting it parse queries. Why bother with Rust anyway? C ticks off the same bullets and the compiler doesn't take a year to run.
  
  In conversation Wednesday, 31-May-2023 18:19:25 JST permalink
  
  matrix07012 :thotpatrol: :cunnyEmpire: likes this.
- CrunkLord420 (crunklord420@rdrama.cc)'s status on Wednesday, 31-May-2023 18:19:26 JST CrunkLord420
  in reply to
  @p @lain @graf @mint @Moon security isn't the reason I want it rewritten in Rust. It's because I refuse to learn Elixir and fundamentally speaking it will always be inferior to a non-VM, non-GC'd language.
  
  Mitra has the right idea just going ahead and using Actix. Actix/Tokio will absolutely annihilate whatever routing framework Pleroma uses (Phoenix?) .
  
  In conversation Wednesday, 31-May-2023 18:19:26 JST permalink
- pistolero :thispersondoesnotexist: (p@freespeechextremist.com)'s status on Wednesday, 31-May-2023 18:19:27 JST pistolero :thispersondoesnotexist:
  in reply to
  @crunklord420 @Moon @graf @lain @mint Didn't KF's most recent token exfiltration happen through a chat widget written in Rust?
  
  Rust is the worst for "all this shit would be better if you people adopted my preferred silver bullet". You hand a browser text headers, you could do it in brainfuck. Rust is shit, won't fix a damn thing.
  
  In conversation Wednesday, 31-May-2023 18:19:27 JST permalink
- CrunkLord420 (crunklord420@rdrama.cc)'s status on Wednesday, 31-May-2023 18:21:20 JST CrunkLord420
  in reply to
  @loathsome @p @lain @graf @mint @Moon there's two answers to this. The real world answer and the academic answer. In the real world the idea of copying large arrays over and over instead of using state and mutability is totally orthogonal to writing performant code.
  
  As for the academic answer I'll just post this meme: https://youtu.be/iSmkqocn0oQ?t=30
  In conversation Wednesday, 31-May-2023 18:21:20 JST permalink
  Attachments
  1. Simon Peyton Jones - Haskell is useless
    
    from bunidanoable
    
    Simon Peyton Jones talking about the future of programming languages
- loathsome (loathsome@petrolkorps.cc)'s status on Wednesday, 31-May-2023 18:21:20 JST loathsome
  in reply to
  https://elixir-lang.org/getting-started/enumerables-and-streams.html#eager-vs-lazy
  
  Have a look at that. I don't have to pass around large arrays, I can just pass around lazy iterators. But I'll forgive you for not knowing this, since you admit to actively not wanting to learn the thing you are talking about.
  
  And obviously making copies of small chunks of information has no meaningful impact on performance.
  
  At the end of the day, languages are tools. You choose the right tool for the job. Elixir is designed to build concurrent, fault-tolerant code and it's absolutely great at this. It can be used for other things as well, but I wouldn't build an operating system with it. It's "good enough" for it's intended purpose. You can obviously be an autistic faggot that thinks any abstraction away from binary code is "orthogonal" to how computers work, but that is akin to only having a hammer and seeing all problems as nails.
  In conversation Wednesday, 31-May-2023 18:21:20 JST permalink
  Attachments
  1. No result found on File_thumbnail lookup.
    
    Enumerables and Streams
    
    Website for Elixir
  matrix07012 :thotpatrol: :cunnyEmpire: likes this.
- CrunkLord420 (crunklord420@rdrama.cc)'s status on Wednesday, 31-May-2023 18:21:21 JST CrunkLord420
  in reply to
  @p @lain @graf @mint @Moon Pleroma is I/O bound partially because it was written in a language specifically designed to be incredibly slow and wasteful when it comes to reuse of memory.
  
  It cannot be repeated enough. Functional programming languages are totally orthogonal to how computers actually work and you can never take advantage of the properties of a computer if you view how a computer actually operates as a flaw that requires a rube goldberg machine to pretend doesn't exist. But of course, never will cease to exist because ultimately you are trying to get a computer to do a thing, not just turn on and get warm (as the intended purpose of functional programming languages).
  
  In conversation Wednesday, 31-May-2023 18:21:21 JST permalink
- loathsome (loathsome@petrolkorps.cc)'s status on Wednesday, 31-May-2023 18:21:21 JST loathsome
  in reply to
  Lambda calculus is Turing complete. I have no idea how functional programming is "totally orthogonal" to how computers actually work. Feel free to explain.
  
  In conversation Wednesday, 31-May-2023 18:21:21 JST permalink
- ICScaryThings (icst@clubcyberia.co)'s status on Friday, 02-Jun-2023 03:33:32 JST ICScaryThings
  in reply to
  @i @p @lain @graf @crunklord420 @mint @Moon i know all the instances mainly get bottlenecked by db writes, but what I don't understand is why? Is it because they don't cache enough in RAM before writing to disk or is it just structured inefficiently so that it has to update more stuff than it should? I feel like something must be missing because a lot of other software out there that uses postgresql doesn't seem to have these problems.
  
  In conversation Friday, 02-Jun-2023 03:33:32 JST permalink
- :blank: (i@declin.eu)'s status on Friday, 02-Jun-2023 03:33:33 JST :blank:
  in reply to
  @crunklord420 @p @lain @graf @mint @Moon phoenix/cowboy would be the equivalents, they're not bad, but pleroma is absolutely gimped by postgresql not keeping up with the buckets of json
  
  In conversation Friday, 02-Jun-2023 03:33:33 JST permalink
- pistolero :thispersondoesnotexist: (p@freespeechextremist.com)'s status on Friday, 02-Jun-2023 03:33:53 JST pistolero :thispersondoesnotexist:
  in reply to
  @icst @i @Moon @crunklord420 @graf @lain @mint
  
  > i know all the instances mainly get bottlenecked by db writes, but what I don't understand is why?
  
  Reads and writes. Part of it is the massive volume of really messy data (messy data always adds overhead but is usually unavoidable in big, distributed, heterogeneous networks), part of it is the architectural decision to keep objects around in their original forms instead of losing data like Masto does.
  
  In conversation Friday, 02-Jun-2023 03:33:53 JST permalink