lemmy.fmhy.ml is gone [update from the team]

Blaze@sh.itjust.works to Fediverse@lemmy.world – 498 points –
very.bignutty.xyz

An update:

  • fmhy.ml is gone, due to the ongoing fiasco with mali government taking all their .ml domains back
  • As such, lemmy.fmhy.ml is also gone, we are currently exploring ways to refederate (or somehow restart federation entirely) without breaking anything substantial
  • We have backups, so don't worry about data loss (you can view them on other instances anyway)

Currently, we have fmhy.net and are exploring options to somehow migrate, thank you for your patience.

175

You are viewing a single comment
View all comments

WIll this also affect all other .ml domains? Or is this some anti-piracy thing? (I don't know fmhy, but from the name I guess it's about piracy.)

It seems to be Mali just wanting their domains back, in which case it's uncertain times for all .ml domains.

rip lemmy.ml

Good thing join-lemmy is safely tucked away in a .org domain.

This is extremely bad timing for Lemmy (if it ends up happening), but also a good example of how federation makes the entire social media landscape more robust. Had this happened to a centralized service it would be devastating.

Not really. Most centralized services are accessible via multiple domains, e.g. for different countries. This would just disable one of them, but users could still use another to log into their accounts. For the Fediverse it "disables" an entire instance, cuts it off from federation and locks out users.

Lets not put a positive spin on a situation that exposes a weakness of the current system. The federation protocol needs to be able to handle these things gracefully, like propagating domain changes and migrating accounts between instances!

I'm now wondering what happens if the Mali government (or someone else) begins using those domains with their own lemmy instance, potentially with malicious content.

Would the instances they've federated with begin ingesting and serving that content automatically? Or would that be blocked due to key mismatch?

Afaik it is all connected to the domain name, so they could definitely start to impersonate any .ml instance. Other instances could detect that the signing key for federation messages changed, but that's about it. Their admins would probably have to block/defederate them manually.

I think they need the private key for the https certificate to do that

If it was always going to happen, now isn't really a bad time. Sure, a month ago would have been better, but people still haven't been here that long. If I wind up needing to migrate, and lose my current account, oh well. No big loss. I imagine others feel similar.

I was frustrated with the outage yesterday and created a new account on a different instance so I could still browse. Couple hours later I had all my subscriptions filled out and the experience is almost identical to my first account.

lemmy.ml is still up as of right now. Possibly they contracted a subscription to the domain name to keep it up. They had to do something to retain it otherwise the site would be unreachable. If lemmy.ml does have to change names it will be a hassle since I've got a good number of community subscriptions there.

This wouldn't happen to an instance with a regularly subscribed domain name. Problem is the .ml domains were free and the associated country decided to claim them back. The risk of using a free top level domain is something that should have been considered. I don't think it's worth the risk versus the cost savings considering how difficult it is to migrate a Lemmy instance.

It's just the domain, though. That's not a big deal to change.

Federation connections are by domain name, so .. it is a big deal

Right. This will basically make nearly every /c live in .world as all of the .ml /c's go defunct. That, or Beehaw, which is walled off from everyone else.

(Side note.. my work's firewalls block everything *.ml -- and that's the only thing that saved me from creating my account there)

deploying the fediverse instances-instance communication on top of a mesh-net like yggdrasil, using their addresses as domain names, may be a quick fix without having to change the paradigm

From that point of view, yes. That'd mess things up, you're right. But from my understanding, they won't lose any data, accounts will remain, as well as subscriptions that lemmy.ml users have. Or am I wrong?

The problem is, if they don't have access to their original .ml domain, their accounts are still tied to it. That means if they try to interact, such as subscribing to a community, when the data for that action tries to be sent back (such as updates) it'll go to the .ml domain, which they wouldn't receive.

Lemmy doesn't have a built in way to just change the domain name, or really any of the ActivityPub services AFAIK. You'd have to either really do some hacky stuff to get around it (which could result in unknown issues down the line) or reset everything.

Most of the hacky ways around it involve retaining ownership of the old domain and leaving it up indefinitely as a pointer to the new location. If your domain is taken from you though there is not much you can do.

Seriously dumb to have used this TLD considering there are a ton of choices these days.

Oh, it's more complex than I expected. Thanks for explaining. I was wrong.

The instance is known by its domain name in the federation network. If that domain name changes, it's like starting a new instance from scratch.

Sounds like a complicated project to migrate communities and posts and users to a new instance without breaking something.

Currently, activitypub identity is tied to domain name. Mastodon support migration as long as the old domain is still up during the migration process, but AFAIK Lemmy doesn't even have a process to migrate an instance to a new domain yet.

Someone should tell Lemmy devs and send them a crate of coffee because it'll be a race to implement domain migration before all .ml domains got shut down.

Shall I make an account in another instance?

Never hurts. Could be a good opportunity to look around the threadiverse and see if you find anything interesting.

However, as it only affects the domain, I expect the Lemmy developers will manage to migrate user data to the new domain should lemmy.ml go down. So your account won't just disappear, but it might go down for a while. It might also affect communities hosted on .ml domains, as followers from other instances will not have the correct path any more.

Yeah, they are actively working on functionality to migrate user accounts and other data between instances, so that they can use that functionality to migrate everything on an instance to another instance.

Since migrating data affects all the replicated data on other instances as well, I guess when they migrate lemmy.ml somewhere else, all of Lemmy will be down for a day or two, being just overloaded with all the migration stuff.

I've migrated from fmhy to feddit.uk, luckily my subscriptions were on a cached web page soon was able to manually re-subscribe.

Nope. Domains don't store data. They can change domain and keep all the data.

Unfortunately, no.

Currently, activitypub identity is tied to domain name. While mastodon support migration as long as the old domain is still up during the migration process, AFAIK Lemmy doesn't even have a process to migrate an instance to a new domain yet.

So basically, if you switch your instance domain, you'll mess up all your federation network, unless Lemmy devs implement a solution soon.

I understand it as the Mali government is taking back all the domains after a subletting contract ran out. A lot of sensitive emails that should go to .mil (US military) has been typo-sent to .ml-addresses instead. Here's some more reading.

(I am very tired here and might have misunderstood everything, please correct me if I am wrong)

Perhaps the military should have a system in place to not allow emails to be sent outside of very specific TLDs if it's that sensitive? And perhaps have an automated contact book, instead of relying on someone typing out the to: address manually to be able to make that mistake in the first place?

Seems like some very basic security measures for something so serious.

Internally they do block that but the problem are people outside the network sending something to a .mil address and mistyping.

For most situations, there is a global address list that members can use. There are instances where emails need to be sent outside of the .mil domain though, such as to other government agencies that use a .gov, or to contractors on commercial domains, as well as to partner nations that will be on their own countries' domains.

Yeah that's super easy to integrate. I used to work in cyber security for a bank and even I was only allowed to send to internal domains initially. I had to file for exceptions for contractors and vendors and stuff.