i started on lemmy.ml but it’s been down for a day or so, cant access anything I did over there, so I started over here too. I’m probably wrong but it seems, from what others have said (over the last few days I’ve been trying out this federation platform), an account’s activity/comments are tied to the sql database on the server - so if the server goes poof, so does your account.
If you and the majority of interested parties here would argue that this “user’s first random host choice, with no backup” is a part of a return to the good old days, this project will never shake any initial bad assumptions. You’ll be doing stupid things for nostalgic reasons.
What was dealt with back in the early days is often completely unnecessary now, and deciding to revive those issues for no other reason than precedence is perfectly analogous to shooting yourselves in the foot. I’ll grant that if real-time/hourly/daily/weekly duplication of accounts to create a recoverable backup is unachievable or unreasonably demanding, fair enough. But if it’s something that could realistically be implemented, and you’d argue it’s not valuable because “if they care so much they should host their own”? Then you’re being downright intellectually dishonest.
There’s a massive gap between any given “Fediverse” host and the global hosting giants when it comes to how much trust a user should reasonably have in their host not suddenly going offline, resulting in a loss of all account information. Other trust issues are another topic, but account persistence should be an obvious and inarguably night and day difference. Spreading that information, or at least a recent duplicate of that information, across two hosts (or more, though not likely necessary) would make that information far more resilient. How could you possibly argue against this?
undefined> What was dealt with back in the early days is often completely unnecessary now, and deciding to revive those issues for no other reason than precedence is perfectly analogous to shooting yourselves in the foot
The reason is to decentralize. Right now there’s like 3 main email hosts. There’s benefits to that, but also downsides. So until there’s some company that runs a lemmy instance (not totally unrealistic), it’s always going to be “some guy’s server”. But even that doesn’t guarantee they’ll be around forever.
I’ll grant that if real-time/hourly/daily/weekly duplication of accounts to create a recoverable backup is unachievable or unreasonably demanding, fair enough.
It’s not about backups. If google stops its email service, you can have all the backups you want but you’ll have lost your @gmail.com email. Same with your lemmy account, if the host goes away you’ve lost your account. It’s really no difference. The actual posts will live on forever because they’ve been federated.
Could/should there be a system to “back up/mirror/migrate” your account to another instance?
Perhaps we could agree on a main instance to back up every account?
If this were to be made, it should maybe have an opt-out? Idk, but it should definitely not be an opt-in thing that misses a lot off clueless new users.
I’m pondering a (strong) suggestion to make a global namespace happen (as it should have been done from the beginning). It would not actually be too difficult. It would mean that communities/posts/users could be handled independently from the domain, thereby effectively virtualising domains, and putting communities/users into a distributed directory. All of it can be backward-compatible and transparent to the user. This would help solve several structural problems at once. First one would be the internal links, then possibility of server migration (without breaking links), resilience against loss of domains/servers (easy use federation as backup) and rogue-going server admins …
I have seen this in so many places now that people come like “this is just not the way you do such a thing”, so I’m thinking about making it a lenghty explanatory post first, for comments to be made, because i might have missed some points and the github discussion board is unsuitable for casual human interaction. Will have to read up on ActivityPub before i proceed … and this is making me a bit sleepless atm ;-)
i started on lemmy.ml but it’s been down for a day or so, cant access anything I did over there, so I started over here too. I’m probably wrong but it seems, from what others have said (over the last few days I’ve been trying out this federation platform), an account’s activity/comments are tied to the sql database on the server - so if the server goes poof, so does your account.
That is a ridiculous way to handle accounts.
Do you feel the same way about email and your email account?
I don’t have my email with a random guy’s server. A lot of new users here do.
But yeah, if my email went down and I had no backup, it would fucking suck. If it was a concern, I’d be wanting a backup system.
undefined> I don’t have my email with a random guy’s server
You kind of do though. Especially back in the early days of email
The only system is hosting your own, which is possible here.
What a ridiculous reply.
I do not.
If you and the majority of interested parties here would argue that this “user’s first random host choice, with no backup” is a part of a return to the good old days, this project will never shake any initial bad assumptions. You’ll be doing stupid things for nostalgic reasons.
What was dealt with back in the early days is often completely unnecessary now, and deciding to revive those issues for no other reason than precedence is perfectly analogous to shooting yourselves in the foot. I’ll grant that if real-time/hourly/daily/weekly duplication of accounts to create a recoverable backup is unachievable or unreasonably demanding, fair enough. But if it’s something that could realistically be implemented, and you’d argue it’s not valuable because “if they care so much they should host their own”? Then you’re being downright intellectually dishonest.
There’s a massive gap between any given “Fediverse” host and the global hosting giants when it comes to how much trust a user should reasonably have in their host not suddenly going offline, resulting in a loss of all account information. Other trust issues are another topic, but account persistence should be an obvious and inarguably night and day difference. Spreading that information, or at least a recent duplicate of that information, across two hosts (or more, though not likely necessary) would make that information far more resilient. How could you possibly argue against this?
undefined> What was dealt with back in the early days is often completely unnecessary now, and deciding to revive those issues for no other reason than precedence is perfectly analogous to shooting yourselves in the foot
The reason is to decentralize. Right now there’s like 3 main email hosts. There’s benefits to that, but also downsides. So until there’s some company that runs a lemmy instance (not totally unrealistic), it’s always going to be “some guy’s server”. But even that doesn’t guarantee they’ll be around forever.
It’s not about backups. If google stops its email service, you can have all the backups you want but you’ll have lost your @gmail.com email. Same with your lemmy account, if the host goes away you’ve lost your account. It’s really no difference. The actual posts will live on forever because they’ve been federated.
Domains (and their lords) owning communities and user accounts speaks indeed some thing about the mindset of the system’s makers, donnit?
Could/should there be a system to “back up/mirror/migrate” your account to another instance?
Perhaps we could agree on a main instance to back up every account?
If this were to be made, it should maybe have an opt-out? Idk, but it should definitely not be an opt-in thing that misses a lot off clueless new users.
I’m pondering a (strong) suggestion to make a global namespace happen (as it should have been done from the beginning). It would not actually be too difficult. It would mean that communities/posts/users could be handled independently from the domain, thereby effectively virtualising domains, and putting communities/users into a distributed directory. All of it can be backward-compatible and transparent to the user. This would help solve several structural problems at once. First one would be the internal links, then possibility of server migration (without breaking links), resilience against loss of domains/servers (easy use federation as backup) and rogue-going server admins …
I have seen this in so many places now that people come like “this is just not the way you do such a thing”, so I’m thinking about making it a lenghty explanatory post first, for comments to be made, because i might have missed some points and the github discussion board is unsuitable for casual human interaction. Will have to read up on ActivityPub before i proceed … and this is making me a bit sleepless atm ;-)