It’s called a single-point of failure in Engineering.
Funny enough it wasn’t even a technical one but a contractual one.
Maybe there is some kind of lesson here on the risk of delegating critical structural elements to 3rd parties that rent rather than own that which they’re selling …
The issue here isn’t the registrar though right? It’s that the TLD is being repossessed by the government of the country it’s meant to be associated with.
Every country gets to decide how tight of a grip they have on their TLD. Some sell it for some extra income (like Tuvalu) while others hang onto it for government or domestic use only
Yeah, and I’d say going with any of the ones that sell it or leave it free is a risk because you never know when their regime might change and the new one might want more of an official internet presence. Unless there’s a 2nd level domain it’s all under (like co.uk), you should assume they’ll want it back at some point. This could apply more to popular domains that some governments could see as free traffic if they reclaim them.
Not really. When you pay for .us domain you have it for a certain number of years. If the US tried to suddenly yank those back and violate the outstanding contracts for x number of years, there would most likely be lawsuits and an injunction from a federal judge blocking the action until there are hearings, etc. It would be a whole thing. If you simply couldn’t renew your .us domain anymore, that’s something you would know ahead of time and could plan for. It wouldn’t just vanish one day.
A centralized system wouldn’t have this problem since the only reason they can’t just use another domain name is because of refederation. A great example of this happening is piracy websites, which notoriously get shutdown only to pop up five minutes later with a new domain.
This is actually a critical flaw IMO in federated applications as a whole. Not being able to change domain names makes your entire platform (as an instance runner) tightly coupled to the initial decision you make when first setting up the instance.
Even better, join a smaller one to spread out and make use of the federated nature. Right now imagine lemmy.ml and lemmy.world for whatever reason go down. Basically whole Lemmy is kinda fucked because it’s extremely centralized, even though decentralization is one of the points of Lemmy.
We need a better way to advertise what servers to direct people to. Would be nice to circle through a big list of instances to evenly direct new users to
It’s hard to make anything “official” on the fediverse as things are distributed. Who makes the “official” determination for Lemmy? The biggest instances? The devs? Do we hold a vote across instances?
People couldn’t care less if it’s centralized or not. People come for the community, not the tech behind it. Also people are lazy, they will use the easiest thing that comes up. Why should one go to another instance, if the one they are right now works great?
I am not saying that this system is bad. I am just saying that people will always take the easiest option there is.
Well, I’m not blaming the people, really. This is a communication issue, it should be well advertised to do it “correctly” and it should actually be the easiest option.
How I see it is that every decentralized system with people is going to form some sort of centralization unless you actively fight against it.
In Lemmy’s case, new people will check what are the biggest communities and go there, and since there are more people there, it attract even more people. More people, more communities in that instance.
I think any non-regional and non-special TLD is fine. Some have rules associated. I thought .movie had special rules about only lasting for a specific amount of time but it looks like I may have been wrong (not sure where I got the idea and I can’t find anything to back that up). .us you have to be a US citizen for. .dev has the “rule” that it is HTTPS only because *.dev is in the HSTS always-on list by default but that’s not related to the domain itself.
The lawsuit points to a 2021 study (PDF) on the abuse of domains conducted by Interisle Consulting Group, which discovered that those ccTLDs operated by Freenom made up five of the Top Ten TLDs most abused by phishers.
Oof. Also what a bizarre landscape where this comes out in 2021 and the only action on it is a private corporation personally suing them over a year later. Where’s ENISA and EC3?
That’s got to be fake news or satire, right? Right?
Like, how is it even possible to send to .ml instead of .mil? ‘i’ and ‘l’ are not even close enough to fat-finger it on most keyboards, and even if you did, it wouls have to be by using the same finger for both ‘i’ and ‘l’, which means you’d physically register the double tap.
I was using .ml domains for my selfhosted services, since it was just an hobby and I didn’t wanted to invest money on it. Apart from Freenom website being pretty unusable since I have memory, I’ve already had troubles renewing them last year and now they stopped working without any notice nor update from Freenom itself. Finally I decided to move to a payed domain from Infomaniak, since it’s been more than a year I’ve been selfhosting and $10/year is a fair price for me.
But still without those free domains I wouldn’t probably ever started selfhosting, and I guess a lot of other people like me wouldn’t have experimented or spin up their projects if they had to pay for a domain from the beginning. So despite my hate for Freenom I guess I have to thank them and hope someone else (maybe a bit more “professional”) will take its place in the future
The lawsuit points to a 2021 study (PDF) on the abuse of domains conducted by Interisle Consulting Group, which discovered that those ccTLDs operated by Freenom made up five of the Top Ten TLDs most abused by phishers.
Umm… Can we talk about how a private company is suing another private company over something that should be in the interest of the government/general public? Where are our agencies, where is Interpol/Europol or ENISA?
Same with Facebook, YouTube, etc. is that highly unlikely? Well, yeah, but still nonzero. The fediverse offers resiliency in this regard, and no one person has the ability to shut it down. Even if all instances decide to shut down, new instances can still be spun up.
If the communities you like to read and post to are down, then Fediverse is effectively down for you. Thus it doesn’t offer any additional resilience, it’s not a P2P system.
Just because anti-lock brakes fail to work in all scenarios doesn’t mean they’re not still an improvement.
Lemmy is still up for most people. That is resilience. If you are affected by this outage, then it failed for you in this particular case but that doesn’t mean the mechanisms don’t exist and that they won’t work to your advantage in the future.
True but if you have several interests, hopefully spread over several instances, then there is resilience because if one server crashes, you have at least some other things trucking along.
Would help if users spread out over all the running servers because problem is just a few lemmy servers have all the users. For example the instance I run would be a simple proxy to use for all the content and then would mitigate issues when a big server had problems since just parts of the fediverse would be affected from the users pov.
I feel like communities are the bigger problem here. And not one that’s easily solved.
If users from multiple instances come together in communities, those communities are still centralized on a single server. So if something happens to that server, or if your instance defederates with it, the whole community goes with it.
The alternative would be to have tons of duplicate communities spread over many instances, but that’s a bad user experience.
I think it can continue even without the source server? Like, once I press the Reply button on this comment, it gets saved to my instance (lemmings.world) then it lets all the other instances know, including lemmy.world (where the community is hosted) and slrpnk.net where you are registered.
Now let’s say lemmy.world stops existing, my instance still would let all the other instances it federates with know, meaning you could read my reply on a community that basically no longer exists. Though I’m pretty sure there are downsides to that (like, what if all the mods were from lemmy.world? There’s no admin who can add a new mod).
meaning you could read my reply on a community that basically no longer exists
oh really? does it actually work this way? if lemmy.world dies, can all its communities continue to live on as long as there are lemmy instances out there federated and subscribed?
No. You would only ever be interacting with a snapshot-at-the-time-of-death of the community on your local instance only. It is the home instance of the community that federates all events, not the instance of the originating post/comment/vote/whathaveyou.
Ah, ok. So if lemmy.world dies, but !somecommunity was federated to 2 different other instances, those instances wouldn’t be able to “talk to each other”? They’d just have snapshots that they could locally interact with, but never see anything else? So is the fate of the Lemmyverse a graveyard of communities from dead instances?
There needs to be a way for a person or group to essentially own a tag to enable moderation. It might be one of those rare problems for which a block chain is a good solution, because there would need to be a public ledger showing who is a moderator for a tag at any given moment.
I wonder about this as well – because communities are tied to a specific home instance, that instance going down affects that community, potentially killing it. Something more akin to hashtags/tags/labels wouldn’t be tied to an instance so they would be more robust, though you’d lose the moderation of a community and just have a firehose of posts/comments…
Wow, you’re right. We really need to bring back something like USENET, where newsgroups (their “communities”) weren’t tied to a specific server. We could almost just resurrect NNTP, although the handling of images (and binary data more generally) probably needs some tweaking.
no need to resurrect it, usenet still exists and has a bit of discussion traffic (and a lot of binary traffic) but we just need to get users to swap over. course there needs to be some decent mobile apps made as well.
Does that really scale though? The load on a server is not dependent on the number of users, but on the number of communities from other server that the sum of user is subscribing to.
Which means if you have a server for 100 users, you still need to pay for the 1000s giant communities that those users are subscribing to, as they are being copied over in your server.
So if you have a few mega server like Lemmy.world, they each pay say 10000£ in hosting a month (number taken out of my hat), which is fine because they have as many users that can contribute to it financially ( via donations, ads etc.). But small servers won’t be able to support that load and will ultimately close.
That sounds like a design flaw if you ask me but i did not see anyone mentioning it so maybe i’m misunderstanding.
No its not really as bad as that at all. The disk space is linear in that way but disk space is cheap. All the rest is not taxed heavily by federation. Do the big costs like CPU dont scale up like that.
At this stage in the game, I’m not even sure how to evaluate the trustworthiness of instances. Which also applies to the one I’m currently on. I’d like to assume everything is good, but admins do have power that can be abused, like visibility of IP addresses, access to accounts, access to passwords (reusing passwords is bad but especially don’t do it here and certainly don’t use the same password for your email associated with your account).
Facebook abused those powers (zuck even bragged about being able to see everyone’s passwords, emails, private messages, pictures), so did Reddit (though more with shadow banning or quietly removing/restoring posts).
Fediverse instances are just run by random people as far as I can tell. I’m sure there’s some that should absolutely be avoided and I’m sure that there’s some that are perfectly fine. But I don’t have a clue how to determine which list about specific instance is in, otherwise I’d love to join someone’s small instance.
Edit: oh and that only goes into whether the admin is acting in good faith or intends to be abusive. Then there’s the question of whether the admin is competent enough to run a server without it getting pwnt and giving others access to that same information and capabilities.
You are correct. A lot of the internet is built on trust. This is no exception. I suggest having an account in more than one instance so that you are not too vested into 1 place.
the problem is most users fear that if they choose a small instance, that it goes down random more likely and their account and everything else is gone. if you choose a bigger instance it feels less likely that the admin of the instance just says fuck it and kills the server random for whatever reason.
as long accounts can’t be easy transfered and are maybe even safe somehow without their instance, people will choose the instance that feels the most secure to them. and when i looked at the available instances… most looked not really long term secure. most did look like they are random ideas of people and they could vanish any second into the void. so i as an example did choose lemmy.world. seemed the most safe option with the best features (nsfw allowed, a lot of users and a big instance)
I understand the logic but its actually backwards. A small instance like mine is easily paid for totally out my own pocket and requires no outside funding or maintenance because I can do everything. If too few people donate to major instances then the costs starts to run away from the owners. In some ways becoming too large is a problem.
i understand that, but think about it - its a random instance from a random stranger on the internet. you don’t know that person, and don’t know if he is actually serious interested in that project of running that instance… or if he will shut it down maybe a few day, weeks or months in the future.
and you can’t really backup your account and load it somewhere else, so if this happens everything you saved and do is GONE. thats a huge risk if you value your account and contribution to communitys.
so it doesn’t really matters to me if smaller instances are not expensive etc… thats not what fears people (there are still ways to spread users along more instances but more even). its the suddenly vanishing without warning that scares people.
i had this often enough with similiar other projects where i created a account on such a small community / instance, was really active… and suddenly it was just gone from one second to the next without warning. everything gone. admin didn’t told anyone about it… was just gone into thin air.
so it feels safer to go to instances who are more “trustworthy” in the longterm security of a stable operation.
if lemmy would support export of accounts maybe ever month once or something… that would change things. also allow spoofing of stuff, but it would help with vanishing instances and people would feel safer on smaller more unknown instances.
“i understand that, but think about it - its a random instance from a random stranger on the internet. you don’t know that person, and don’t know if he is actually serious interested in that project of running that instance… or if he will shut it down maybe a few day, weeks or months in the future.”
Have to be honest with you, that is how all yhe instances started including lemmy.world.
“so it feels safer to go to instances who are more “trustworthy” in the longterm security of a stable operation.”
There is no metric by which to know this yet as lemmy is new. Its not like there are 5 servers that are 10 years old and al the rest are just starting up. Just how it is.
Have to be honest with you, that is how all yhe instances started including lemmy.world.
but now they have enough reputation & users to make them feel like the safest option
There is no metric by which to know this yet as lemmy is new. Its not like there are 5 servers that are 10 years old and al the rest are just starting up. Just how it is.
compared with random instances with 2-3 users or so, a instance who is there since the beginning / relative long compared to other is safer feeling tho.
i’m so worried about this topic, that i even think about maybe setting up my own instance just to keep my accounts etc safe & from vanishing.
I feel like you have missed the points im my previous comments but if you just want to feel safer because in your heart of hearts this instance or that instance just feels safer then go for it.
My advice does not change. Make a backup account on another instance to avoid being burned. If you dont want to, then its now on you.
On a small instance, you have greater opportunities to take action to positively support that instance. You can make friends with the administrator, volunteer to become an administrator yourself, donate cash to offset running costs, lodge helpful reports, welcome new users, etc…
agreed, but i’m already moderating a community with 1,3k members elsewhere and have to do a lot of work daily there (posting content for the members who wait for it daily). also i currently start to build one up on lemmy.world that also takes time from my day. i don’t really have time in my daily activity to additonally do stuff which involve moderation or managing of such things like a server instance.
don’t understand me wrong, i agree with what you say and its logical and smart to do it. but its always depending on the situation of each user. in my situation, its the best thing to go to a big instance.
My exact same thought process and why I’m here on lemmy.world as well. Once they get the server setup process more streamlined (hopefully dockerized) I’ll probably setup my own private use server, but until I get around to that project I wanted to pick one that didn’t seem like it would vanish once the guy hosting it started getting those hosting bills.
A week ago I literally read articles about how .ml was switching to the (Russian-influenced) Mali government in a week, and did not even think about how lemmy.ml would be affected
I think that’s different because the .ml domain apparently was being given away for free by a registrar that wasn’t responding to abuse complaints, and thus was being heavily abused.
…but if not, then holy shit what a mistake it was to register [email protected] as my primary email address.
the argument for .ml domain has always been absurd to begin with. So it’s free but the price you pay is that it’s being run by Mali. I’d just drop 8$/year tbh, that’s not a hill you want to die for. Also you harm your project by being SEO punished for using spam-associated TLDs like this. One of the reasons original Lemmy took so long to adopt until Reddit’s API drama. Pretty dumb ngl.
Anyone know how companies get the rights to domains to sell in the first place? Do they literally submit a list of all domains to ICANN or something? Sorry if this is a stupid question, I just never understood how any of this really works.
ICANN hands out top-level domains (TLDs - such as .com, .org and .ml), either to organisations or government agencies. They, in turn, hand out secondary domains to companies or regional organisations. For example, the TLD .jp belongs to the Japanese government and is operated by an agency called Japan Registry Services. In turn, it hand out the .tokyo.jp secondary domain to the Tokyo Metropolitan government. They, in turn, manage domains for various departments, wards, etc.
But individuals and businesses in Tokyo can also use the .tokyo TLD, which is owned by a private company called GMO Internet Group. And of course anyone can use .com or .org, although you may have tp pay a pretty big fee.
Because it means losing access to the content unless the operators go through the ardous process of moving to another domain, whereas with the matrix protocol the content would remain perfectly available and the only thing that happens when a server has domain issues is that the accounts and specific room alias become unusable.
XMPP has the same issue because it also relies on one central server to host a room, whereas with matrix ALL involved servers replicate the room which means that there is no central server to go down, which is just objectively better for things like chats and forums.
Technically instances do actually duplicate the communities. Right now I’m responding to a comment cached on the database of lemmy.thesanewriter.com, and I would retain that ability even if your instance defederated from mine or went down forever.
With activitypub all involved servers also replicate the content so I'm not sure what distinction you're trying to make. That's why we can still see all the communities, posts, and comments on the servers that are still online.
that is not entirely true from a lemmy perspective. When an instance subscribes to a community, the remote instance gets the last 20 or so posts, as well as subscribes to all new posts from then on. IT has a local copy of that community. What it doesn’t have is any of the embedded media.
AFAIK, this is similar to how matrix works too. I do not know if this is a lemmy implementation choice, or a AP standard?
Edit: haha, i just saw that i am the 3rd person to say the same thing. oops!
I really appreciate the links! So it looks like matrix has mainly been focused on the chat room side of things and another downside is that you need some sort of translation layer to connect a matrix thing to an activity pub thing. Definitely gives a lot to think about. I’ve been trying to decide where I should donate my time to doing dev work at but the answer appears to be “it’s complicated”
Add comment