After a year online the free speech-focused instance ‘Burggit’ is shutting down. Among other motivations, the admins point to grievances with the Lemmy software as one of the main reasons for shutting down the instance. In a first post asking about migrating to Sharkey, one of the admins states:

This Lemmy instance is much harder to maintain due to the fact that I can’t tell what images get uploaded here, which means anyone can use this as a free image host for illegal shit, and the fact that there’s no user list that I can easily see. Moderation tools are nonexistent on here. It also eats up storage like crazy due to the fact that it rapidly caches images from scraped URLs and the few remaining instances that we still federate with. The software is downright frustrating to work with, and It feels less rewarding overall putting effort into this instance because it feels like we’re so isolated.

A few weeks later, in the post announcing that Burggit was shutting down, another admin says the same:

The amount of hoops that burger has to go to in order to bring you this site is ridiculous. To give you an idea of how bad this software is, there’s no easy way to check all the images uploaded to the site (such as through private messages). When the obvious concern of potential illegal imagery is brought up to lemmy devs, they shrug and say to plug in an expensive AI image checker to scan for illegal imagery. That response genuinely has me thinking that this is by design, and they want it to be like this. We can’t even easily look at the list of registered users without looking through the DB, absolute insanity.

The other thing is there’s no real way to manage storage properly in Lemmy, the storage caches every image ever uploaded to any instance forever.

Also the software is constantly breaking.

They also say that Kbin has many of the same problems, so I’m just curious to know if the admins of bigger Lemmy & Kbin instances feel the same way about these software.

  • nutomic@lemmy.mlEnglish
    5·
    5 months ago

    I looked at some of the pull requests and most of them seem very small, only changing a couple of lines. Still impressive but not really comparable to implementing a new feature in Lemmy. For that we need to make changes to various different parts of the code (database, federation, api, js library, frontend), then test it and pass code review. All that takes a lot of work because we need to ensure that existing functionality doesnt break. In this way a web server like Lemmy has much higher standards because there should be no bugs at all. If your AI project has some bugs, users can easily roll back their local install to an earlier version.

    Youre right about lemmy-ui, unfortunately it doesnt have enough contributors. I dont know why that is, you’d think a project written in a popular language like Typescript would easily find contributors.

    • maegul (he/they)@lemmy.mlEnglish
      3·
      5 months ago

      Youre right about lemmy-ui, unfortunately it doesnt have enough contributors. I dont know why that is, you’d think a project written in a popular language like Typescript would easily find contributors.

      Random thoughts:

      • Is it obvious enough that one can contribute to the UI separately from the backend and that it’s a Typescript SPA style UI?
        • If not, maybe a bit of a “dev recruitment campaign” could help … let people people know and what sorts of issues could really do with new contributors lending a hand? Maybe even a bit of a “Inferno isn’t that different from all of the other SPA frameworks/libraries spiel?”
      • Is the use of Inferno as oppose to one of the big 3 React/Vue/Svelte a repellent? (perhaps a downside to the “diversity” of frontend frameworks?)
      • Are would-be UI contributors more inclined to make their own front-end or app than contribute to the default webUI?

      More generally:

      • Would a server side rendered webUI be welcome?
        • Then the contributions would mainly be on templates and their “simpler” logic, which might be more attractive or easier to get started on?
        • Plus, it might be more efficient? The current UI feels to me like it would suit server side rendering well.
        • Is this where the new leptos UI is heading … more server side rendering (I don’t know much about leptos)
      • Do you have a sense of usage numbers for the different apps and frontends? Obviously you only run lemmy.ml, but do you have a sense of how much the front-end gets hit versus the API directly?
        • I ask, because If the default WebUI is really the main interface, then it makes sense to try to organise some more contributors (It’s certainly my main, nearly exclusive interface, as much as I’ve like some of the alt front ends or apps)
    • h3ndrik@feddit.deEnglish
      11·
      5 months ago

      Yeah, you’re not doing it right. On Github you have to click on “Insights”. And alike Lemmy which is split into two parts, llama.cpp also has a backend called ggml that does the (tensor) maths. Combined, the git stats are as following for the last four weeks:

      • Lemmy (+UI) 207 files changed, +7,841 additions and -6,472 deletions
      • llama.cpp (+ggml) 707 files changed, +157,754 additions and -95,611 deletions.

      So they definitely touch a lot more code regularly. Whichever PRs you clicked on, they added 50 times as much new lines of code in the same timeframe. And coding things like that is maths heavy and you also need to read the scientific papers and implement the maths. And they did quite some maths themselves and contributed their quanitzation techniques and benchmarked and studied them in addition to the coding. I’m really impressed by the guy. And he seems nice and attracted quite some contributors with his excellent and fast software. Reviews and comments their ideas and integrates them fast. And now it’s a flourishing project that leads in its field. And the project isn’t even that old…

      I get it. Software development isn’t that easy. Especially the ‘touching different parts of the code’ is something I don’t really like. I mean it is like it is. And having architectural patterns like this is fairly common (logic, database, UI) and you have like 2 models of the data, one for federation and then the internal representation. I’m not that familiar with the Rust frameworks and how cumbersome it is to deal with them. With the correct database abstraction toolkit and other frameworks it gets better and you can often tie the stuff together. Also helps with the bugs. If it’s really bad, maybe the architecture isn’t optimal. Or the chosen frameworks suck. Other than that it’s the job of a programmer to tie those aspects together, deal with the complexity and combine it into a working product.

      I’m not even sure if you can assure that Lemmy has no bugs… I mean unit tests, integration tests and reviews won’t cut it with distributed or federated software, right? I mean you’d need to roll out a small cloud of instances and do end to end tests, check if everything federates and if there are performance regressions… I’m not sure where Lemmy is regarding this. I occasionally observe when something big happens like federation breaking.

      Sure. And UI programming is also something that is not really fun to me. I’m also not sure why it hasn’t more contributors. Maybe the atmosphere isn’t that welcoming to new people. Or the userbase in total is just too small. I mean fediverse observer reports like 50k Lemmy users, and that’s not that much people if we’re talking about the subset of people who learned programming and have the spare time to contribute. Maybe it’s too interlinked with the rest of the code or not documented enough. I’d say it’s probably not that attractive to get involved because it’s mainly small bugfixes that can be implemented without also getting involved with the rest of the project. And apart from drive-by pull requests, people usually have some bigger vision when they join a project.