PriorProject @ PriorProject @lemmy.world

Just an explorer in the threadiverse.

---

Read more

Posts

9

Comments

266

Joined

2 yr. ago

Not to be glib, but the way you fix >200 communities is the same way you fix the first one, and then do it >199 more times.

It's broken and it sucks. People have articulated the hacky workarounds that sometimes work for them, there's no bulk version of that process that I'm aware of. You either ignore it, bang on it one community at a time until it's fixed, or become a programmer and fix it for everyone (or script the hacky workaround).

I don't have a link handy, but I saw this reported before and when someone went to lemmynsfw to check the posts, they actually had been properly marked as nsfw.

My unsubstantiated theory is that federation is working pretty crappily right now and so lots of federation messages get dropped. When a new post or comment gets dropped you don't notice because... well... you don't see anything. But when a post makes it through and an edit gets dropped, that can be more visible.

So likely what's happening is that on some small percent but medium absolute number of lemmynsfw posts, someone makes an honest mistake and fails to tag it. They either notice themselves and fix it, or a mod asks them to and they comply... so on lemmynsfw it looks right. But the federation message with the edit gets dropped by your instance, and for you and others viewing from there it forever remains sfw in error, even though all the right steps were taken. I've also heard mods discussing this happening with post removal moderation actions, the post gets removed on the community's home instance, but the federation message containing the removal gets dropped by some big instance and the post blows up there anyway where the mods are actually unable to shut it down.

I don't think there's anything anyone can do about this other than maybe the lemmynsfw admins tuning their federation worker counts (though they may have already and that may no longer be needed in recent Lemmy versions), or the Lemmy devs working on federation scalability so a larger percentages of federation messages get reliably delivered.

I thought that was the first rule of rendering web content? Or was it protocol parsers?

I remember, it was first rule of video game character creation screens:

How are you accessing lemmy.world? Via desktop browser? Try clearing your cache. Via mobile on Android, clear android's app data for your app. Also try an explicit logout to ensure your session is cleanly torn down on both sides.

Sorry, that password is already in use.

Who's using it? I'll just use that account.

Head scale would be a self-hosted way of doing this as well.

• You'd install headscale publicly accessible on your VPS or port-forwarded server.
• You'd configure your phone and any laptop you travel with using the tailscale apps with the special hidden setting to use your custom control-server.
• Now any apps you want to access yourself but not for the public unauthenticated internet to see, you bind to tailscale/headscale interfaces rather than public interfaces.
• Anything you DO want publicly accessible (for example immich for image sharing to friends who aren't on your tailscale network) you host the normal way by binding to a public interface.

You could also do this with regular tailscale and cut the self-hosted headscale out of the picture.

But by doing this or another private VPN setup, you take the listeners for some of your apps off the internet and reduce your attack-surface. It obviously doesn't help for WordPress or other stuff you actually want to share publicly, but it can give some peace of mind for personal services like bitwarden or Jellyfin.

This isn't a terrible idea, but it's also important to understand single-user and tiny invite-only instances as analogous to "leechers" in the torrenting world. The federation load that an instance instance imposed on other instances depends much more on the number of communities it subscribes to than the number of active users. If a user stops using Lemmy but leaves their instance up, it's generating federation load for no reason.

Tiny instances are inefficient, and while it is desirable for the network to be able to scale to the point where it can reasonably support lots of them anyway, right now federation queues are backed up and messages are frequently getting dropped. Encouraging lots MORE tiny instances is probably not the efficient thing right this second. Rather, we'd want more users joining mid-sized instances that are not overloaded locally and that are making efficient use of the federation load they generate by using it to serve 100-1000 users rather than 1 or 2.

Liability is not binary. There is a qualitative change in risk as you transition from "I subscribed to 100 actively moderated communities that I read and am familiar with" toward "I subscribed to everything there is including the worst of the worst and I didn't realize I was doing so and don't look at the results".

Also, moderation activities federate. So even if a rogue poster does "contaminate" the actively moderated communities on a well-admin'ed instance... when those mods and admins delete the offending material they'll automatically cleanup your instance as well. As a result, it's the creepy crawly communities that don't clean up or don't want to clean up that generate the lion's share of risk.

Is it 100% safe to sub to well-moderated communities, no. You have to know your local laws and protect yourself. Do you do yourself favors by running lemmony? Also no. These two statements can be simultaneously true.

The upsides are that you control your defederation list and you're your own admin so you're in control of whether your instance goes down and what it's policies are.

The downsides are:

• Potential privacy leaks. Your all feed is public. If its full of creepy shit and you're the only person in your instance, it's there cause you subscribe to creepy shit.
• You're in control of whether your instance stays up. Security vulnerability gets mass exploited? Your problem.
• Potential hosting liability. Your instance mirrors what you sub and serves it to the public unauthenticated internet. If you subscribe of stuff that's questionably legal in your jurisdiction, that liability can become yours unless you're familiar enough with your laws to know how to protect yourself.
• All the standard self-hosting stuff like cost and hassle.

Folks should not use lemmony to bootstrap their subscription count. It's not that hard to hit lemmyverse.net and just manually sub a bunch of stuff you're actually interested in, or to visit a big instance and browse their all feed unauthenticated.

But if you really want to automate community bootstrapping, lemmony is the worst of the scripts that doit because it defaults to subscribing to EVERYTHING, including all the porn, piracy, and hate communities on the most absent-admin'ed under-modded instances in the lemmyverse. Then your instance will mirror all those questionably legal communities and re-serve them to the public unauthenticated internet, creating hosting liability for you. Not to mention being a bad fediverse citizen and creating massive amounts of federation load on the instances forwarding you posts and comments from 20k communities that you don't read.

These two subscription bootstrapping scripts limit you to top subs by default... So you're more likely to be in well-modded territory and just the number of subs is smaller you you can review them and back out of anything sketchy. Subscriber-bot's docs do a good job of explaining the risks and problems of mass-subscription so you know what you're getting into.

• https://github.com/lflare/lemmy-subscriber-bot
• https://github.com/Fmstrat/lcs

As others have noted, this was an issue that was frequently with Jerboa ~v0.34 or so with Lemmy v0.17.x and was attributed to the buggy websockets API. It initially stopped happening to me once lemmy.world upgraded to Lemmy v0.18.x and I got on Jerboa v0.37 or so... and most reports of it died down as well.

But in the last couple days, I have been seeing it as well. Including today. I'm now running Jerboa v0.40 against lemmy.world with a v0.18.2 build... so it's not websockets anymore it must be a different underlying cause. But I definitely do confirm your report of seeing the wrong post and fixing it via reloading, though I can't explain why. I also use liftoff and voyager via m.lemmy.world and I haven't seen the problem there. I assume it's a local bug in Jerboa this time, rather than a protocol issue.

So anyhoo, you're not alone in seeing this and although folks are correct to say that a similar issue was already fixed... it's possible to see this with fully updated Jerboa. In fact, I think it must be a bug that was introduced somewhere between v0.38 and v0.40.because it stopped happening for a while after things upgraded away from Lemmy v0.17.x and the jerboa versions form around that time.

It has to do with how federation works. There's no federated broadcast to all instances in the threadiverse when a community is created. Each instance has to be taught about each community via a search. I think this was done to prevent broadcast storms due to community creation, though in normal circumstances community creation is not a high frequency activity. I think the world would be better off if community creation broadcast but was rate limited.

At any rate, in the meantime search for communities at lemmyverse.net. It's way WAY better than the native community browser. Then just use the native community browser to find an sub the specific thing you found in lemmyverse.

Connect for Lemmy by Kurononeko? Is that open source, or only available in the android store?

Yeah, I'm addressing the current actual state, not the desired state. The only other thing I can think that might achieve a similar result to deletion is blocking the user. I haven't blocked anyone, but that might hide the reply (it should IMO). Not sure I'd you want to go that drastic in this case.

• Liftoff
  • Animated in feed view
  • Animated in single-post view
  • Animated in image view.
• Jerboa:
  • Static in feed view
  • Static in single-post view
  • Animated in image view
• Voyager PWA in Firefox on Android via m.lemmy.world:
  • Animated in feed view
  • Animated in single-post view
  • Animated in image view
• Mobile PWA in Firefox on Android via lemmy.world, v0.18.2
  • Animated in feed view
  • Animated in single-post view
  • Animated in image view.
• Firefox desktop via lemmy.world, v0.18.2
  • Animated in feed view
  • Animated in single-post view
  • Animated in image view.

Seems like it's working pretty well overall. Jerboa is the outlier, but you CAN see the animation if you click into it.

That said, we don't all need to crowd onto one instance and shouldn't do so. If you're advising new users, I'd point them toward an instance with 100-1000 users. With your month-old active account... this is your home though and if you can be patient I think they'll start to get ahead of the growth as long as people don't keep dog-piling here exclusively.

You've stumped the chump, apologies.

Oh, I see we're talking about your login issues I'm a different post. Ok, good luck with the community takeover if your login problems don't get sorted.

Weird. Did you try a different browser, using a mobile app, or using the password reset process?

I was hit by the mass logouts like many others, but didn't have much trouble getting back in using the techniques you've already tried.

Here's a potentially unpopular opinion... Games that target the Proton API are actually native Linux games. Proton isn't virtualization or emulation, it's just an API that happens to be mostly compatible on both Windows and Linux. Other than the kernel itself, Linux has never had one true API to do anything... there's always more than one option to target (as you note with your Wayland/x11 example, but also pulse, alsa, pipewire, the list is endless). Proton is an API that's available on Linux, and programs that target the Proton API are Linux programs in every way that matters.

The question isn't native vs proton. The question is whether proton is a good API. At the moment, it's an API that offers pretty good cross platform compatibility with windows, which is hugely valuable to developers and they're using Proton for that reason and even testing against it. That's good for us as users and for gaming on Linux.

If Windows evolves their versions of the proton APIs in ways that break compatibility and are difficult to fix, we may find that game devs complain on our behalf to avoid breaking their Linux builds. If Proton begins to suck compared to alternatives, and enough people are playing games on Linux with Proton, devs will organically start to look at other porting options more seriously. But Proton is both a way to kickstart the chicken/egg problem, and itself may just actually be a good API to develop Linux games against.