Hey… that just gave me a small idea… what if we made a “flock” or “herd” of Mastodon servers? The group of servers would all federate with each other, have the same block and allow lists, moderation policy and teams spread throughout them.
When you make an account you can be assigned a random instance name within the flock. If your instance goes down you could still possibly log in using other servers? Main benefit would be spreading server costs and maintenance effort and de-centralized operating, but still keep a centralized feel to it?
Honestly that’s probably the best sort of solution. A group that has some minimum standards of moderation and maintenance/upgrade management plan and just evenly distribute the load as people arrive.
Then as a second phase make it easy to transfer, that way at the point the user gets comfortable they can easily swap to a better* “home” for those that care, for those that don’t, make the server choice be virtually invisible.
If the fediverse ever wants to scale, something like this has to come about. I personally think we need a whole lot of regional servers. For example, we make a cluster of servers by country, so lemmy.us, Lemmy.de, etc. Then, when those servers start to fill to a certain threshold (say 1000 users), we break them out regionally, so lemmy.ne.us, lemmy.se.us, etc. The way servers are assigned would be by selecting your country and region. It shouldn’t be too complex and would simplify the sign up people for a lot of people.
We already see lag in comments and posts with the current load. It could be buggy apps as this is relatively new, but who knows. I am no network engineer, but I would imagine that issue to only get worse as user numbers increase.
If you see lag, you should try using a different instance. LW was noticeably slow during summer 2021, a lot of people moved to other instances due to that
If they have the same people running all of them, how is that different from running a single mastodon server in kubernetes, so that it doesn’t get overloaded?
You’d have different domain names to get people used to the concept. John Doe would sign up, and become [email protected], Jane Doe would sign up and become [email protected]
That is what we have now, but clearly people are averse to making a choice that they are not technically inclined to know how big or small the consequences of that are. My solution is a spitball one with obvious flaws, but essentially it is that the instance is picked randomly out of a group of very closely, if not identically aligned servers.
Hey… that just gave me a small idea… what if we made a “flock” or “herd” of Mastodon servers? The group of servers would all federate with each other, have the same block and allow lists, moderation policy and teams spread throughout them.
When you make an account you can be assigned a random instance name within the flock. If your instance goes down you could still possibly log in using other servers? Main benefit would be spreading server costs and maintenance effort and de-centralized operating, but still keep a centralized feel to it?
Honestly that’s probably the best sort of solution. A group that has some minimum standards of moderation and maintenance/upgrade management plan and just evenly distribute the load as people arrive.
Then as a second phase make it easy to transfer, that way at the point the user gets comfortable they can easily swap to a better* “home” for those that care, for those that don’t, make the server choice be virtually invisible.
i like the idea of a server choice virtually invisible feature!
Let me see how you get instance admins to agree on what to defederate.
Maybe a vote of 75% minimum would be good?
If the fediverse ever wants to scale, something like this has to come about. I personally think we need a whole lot of regional servers. For example, we make a cluster of servers by country, so lemmy.us, Lemmy.de, etc. Then, when those servers start to fill to a certain threshold (say 1000 users), we break them out regionally, so lemmy.ne.us, lemmy.se.us, etc. The way servers are assigned would be by selecting your country and region. It shouldn’t be too complex and would simplify the sign up people for a lot of people.
You’re assuming that all those servers would have the same policies and admins.
As we can see from their recent announcement, the LW team has some specific policies in their Terms of Service that no other instance replicates.
You are thinking about load balancing, but that can be handled by Cloudflare or something else, it’s doesn’t have to be a different instance.
We already see lag in comments and posts with the current load. It could be buggy apps as this is relatively new, but who knows. I am no network engineer, but I would imagine that issue to only get worse as user numbers increase.
If you see lag, you should try using a different instance. LW was noticeably slow during summer 2021, a lot of people moved to other instances due to that
If they have the same people running all of them, how is that different from running a single mastodon server in kubernetes, so that it doesn’t get overloaded?
You’d have different domain names to get people used to the concept. John Doe would sign up, and become [email protected], Jane Doe would sign up and become [email protected]
This is quite unnecessary, it would be simpler if we have a list of the long-running and most stable instances and have the users pick one.
That is what we have now, but clearly people are averse to making a choice that they are not technically inclined to know how big or small the consequences of that are. My solution is a spitball one with obvious flaws, but essentially it is that the instance is picked randomly out of a group of very closely, if not identically aligned servers.
Basically, a single instance
Where?
When you go to comment on a blog, where do you sign up?
Man, it feels like you guys haven’t spoken to a real human in decades…