Lecture intéressante, qui met bien en avant les différents avantages et inconvénients de Lemmy, et la potentielle solution qu’est Sublinks

  • Penguin_Rocket@lemmy.ml
    link
    fedilink
    Français
    arrow-up
    1
    ·
    10 months ago

    Comment vous faites pour lire le post ? Quand je clique dessus je n’obtiens qu’une visualisation d’un fichier json dans Firefox sur ordi, et je dois télécharger un fichier .bin quand je le visualise sur smartphone

    • Camus (il, lui)@jlai.luOP
      link
      fedilink
      Français
      arrow-up
      3
      ·
      10 months ago

      jewy.blog My Love-Hate Relationship With Lemmy – Gavi’s Blog 15–18 minutes Content Warning: This blog post contains discussion of CSAM & harassment on the Fediverse (as well as other nastiness)

      Background
          My own Lemmy (& Fediverse) background
      First, the good
      The Bad
          The Software Itself
          The Developers
          The Community
      The Kiwifarms
      What next? (Sublinks, maybe?)
      

      Background

      So… what is Lemmy exactly? It’s unlikely you know unless you’ve actually used Lemmy before or have an interest in fediverse software development. It’s especially unlikely that you’ve seen a Lemmy post within the Fediverse as although it is built atop ActivityPub, it’s compatibility with other Fediverse software is currently very limited.

      If Mastodon is a Twitter clone then Lemmy is Reddit’s Dolly the Sheep. It’s a software in very early development that’s been around for a few years now (One of it’s first releases on Github was May 6th of 2019) but only relatively recently during the Reddit API debacle did it gain adoption beyond a very small niche. Compared to the rest of the Fediverse it is still very small but it is steadily growing and as seen by other corporate platforms it is unlikely that this will be Reddit’s last controversy.

      It has a multitude of problems, and my hope with this blog post is to highlight them as well as potentially bring attention to some glaring issues I feel are just… completely ignored. My own Lemmy (& Fediverse) background

      I joined Lemmy during the API debacle but my own experience and knowledge of the Fediverse didn’t begin with Lemmy. I have followed the development of the Fediverse and the software around it for years. I would sign up and occasionally check out different Fediverse software, but never really committed to using any until this past summer.

      When I first signed up for a Lemmy instance it was just out of curiosity. I had no desire to have my own instance, provide admin support for other larger instances, or even weigh in on its development. It was just curiosity for curiosity sake. Instantly though, I began to see the potential Lemmy has. In spite of everything, I still see that potential too for such a platform. First, the good

      I think before I really get into the weeds of my criticism with Lemmy it’s probably best to start with a bit of what I actually like about it. Because there are some things I actually do enjoy! I wouldn’t use it or still have my own instance if I didn’t.

      The principle behind the software works really well for a federated system. Instead of a giant single instance with its own singular feed, an instance has communities with their own structure based atop the instance itself. It’s similar to Reddit in that regard. Instead of just having a sitewide singular culture there’s room for more than one community culture to grow on a single instance.

      And in my own opinion, that makes more sense for large scale instances compared to large scale microblogging Fediverse instances. Having a smaller cluster of communities that have their own moderators rather than one giant community as an entire instance feels a lot easier to manage as an instance admin if you have sufficient moderators (and tooling, but we’ll get into that….). Granted, I’ve never been an admin on a TRULY giant Mastodon instance but I imagine it’s very very different. But large Mastodon instances have a particular reputation of being not well moderated. There comes a point where managing an instance of such a giant size just feels impossible to be done correctly. With Lemmy that feels different.

      When I’ve provided admin support for larger Lemmy instances the communities essentially babysit themselves. It’s quite smooth when it actually works as intended!

      I personally enjoy the forum based system for conversation around news and other topics a lot more enjoyable compared to on Mastodon. That’s not to say I dislike Mastodon or the microblogging side of the Fediverse, I just prefer the rest of the Fediverse for friends and casual conversation. Lemmy is where I feel more open to being a bit more serious with what I interact with in the Fediverse. The Bad

      There’s a lot to unpack here, so I’m gonna split it by category. The Software Itself

      I feel very averse to bringing attention to some of the issues currently present with Lemmy as many of these are likely potential attack vectors by bad actors. But at this point, I don’t know how else to point them out.

      A bullet list of the biggest issues with Lemmy’s software that renders it extremely stressful to manage a community on it right now:

      Extremely insufficient moderation tooling.
          The only site wide staff option is admins. There is no granular permissions, you are essentially given the keys and then some.
          There is no granular federation options. Only domain blocks and that’s it.
          Removing communities don’t remove them entirely, content still gets federated in but is hidden to staff only.
          Image caching is forced on every image.
          Extremely insufficient documentation.
          Reports are auto forwarded without any anonymity. And yes, report retaliation bans happen from community mods.
          Reports are sent to both admins and community mods, with no means of forwarding directly to admins. Community mods can snipe reports before admins see them.
          Domain blocks are always publicly visible.
          Mod logs are always publicly visible in the public mod log.
          And a laundry list of more.
      

      The extremely insufficient mod tooling became a big issue rapidly during a series of attacks that occurred over the summer that culminated in CSAM being posted in a large Lemmy community. I was one of the first people outside of its original instance to call attention to it being federated outwards, and had to witness and deal with the fallout. The issue that caused it were pointed out months prior and completely ignored.

      This blog post by a Lemmy user who accidentally uploaded his ID and dealt with the nightmare after describes in great detail the ridiculous steps instance admins need to take to remove images from the backend image server that Lemmy depends on. (as well touches upon the developer behavior aspect I will highlight later.)

      In order to sufficiently purge images you need to have external tools and scripts that the community maintains. There are some instances that utilize local machine learning algorithms to scan images, but most instances are unable to do so because of the GPU power that requires.

      When the CSAM started federating outwards across the network, I had to delete my instances storage buckets from the 24 hours to when the material was identified numerous times. As well, with the scripts provided to do a single removal you still need to view the images to get their URLs or rely on another instance admin who has done so for you. At this point Lemmy admins coordinate in matrix channels over this, and some days it’s worse than others. It is unreasonable to expect instance admins to have to rely on such tools outside of the software itself. Most can’t. The Developers

      Something that I notice said consistently by those who have little experience in Lemmy admin spaces is “why not just contribute then?”

      And the answer people try. And this happens. This unfortunately leads into the next point that is the developer teams behavior. As well, highlighted above in the blog post of that Lemmy user who unfortunately had to deal with devs behavior themselves.

      In the wider Fediverse Lemmy has a pretty bad reputation due to the developer team and did so far before the Reddit exodus. But, the degree as to how little trust most Lemmy instance admins have in the Lemmy development team is often understated. There is a fundamental lack of confidence amongst a majority of Lemmy instance admins towards the lead developers of Lemmy.

      The devs are right: the software does not come with a warranty. But people will leave your software behind if you refuse to engage in dialogue as well as exhibit petty and hostile behavior towards those who actually are attempting to take the time to try and contribute to the project. Above isn’t the only example I have seen, there’s numerous others. This is the developers project yes, but opinionated development takes a different level when you actively refuse to implement features to improve inter-Fediverse compatibility due to a grudge against Mastodons overall dominance in the Fediverse. People don’t contribute because those that do just eventually don’t see it as being worth their time.

      Another aspect of this is that the Lemmy devs run two instances: lemmy.ml & lemmygrad.ml

      I often see lemmygrad.ml in many instance blocklists across the Fediverse…. but you do know that lemmy.ml is just lemmygrad lite right? That’s not hyperbole, the admins and moderators are nearly identical and the moderation style of what is action-ed on is consistent with both. Lemmy.ml is more or less the normie facing flagship instance and lemmygrad.ml is the communist shitposting side. This reason makes it difficult for many Lemmy instance admins to really trust what Lemmy devs have to say as there’s not much way of telling if Lemmy devs are using their lemmygrad shitposting accounts to shit talk other instance admins. (It’s happened before.)

      The biggest piece that broke all confidence in the Lemmy developers amongst many admins including myself is that during the CSAM spam attacks there was complete radio silence. The developers made no statement on the matter. And when Github requests were made to try and propose ideas about how to fix what happened, the developers explicitly stated they didn’t have time to focus on that. No dialogue.

      As well, when a post was made about Sublinks (A project I will touch a bit more on, and am involved in due to the reasons I have highlighted above) the comments that were made by Lemmy’s lead developers were extremely petty. This lessens