I’m fairly new to fediverse/instances, but my understanding that these Instances are federated with each other, so they can show the same posts/communities (depending on their respective filters).
So, why does searching for communities show different results, and even show different subscriber counts for one that do appear across all results?
Is this a case of the servers just needing to sync with each other?
Every server does not contain a record of everything that’s on every other server. That would be way too much, and essentially mean that every server would have to download the entire fediverse, that’s not how it works.
Each server only knows about what’s directly on itself, and what at least one user has interacted with.
Even if my server is federated with yours, my server wont sync posts from a community your server has, until at least one user on my server has subscribed to that community.
This way, each server only needs to contain stuff that the users on it actually want.
As for the subcount, each server is showing how many users on it has subbed, not the global count. For that, you need to go to the host instance.
I don’t think the host instance has a global count, even there they only have data on local users.
Then how does [email protected] have 9.3k subs when viewed from sopuli?
Sopuli only hosts 3.7k accounts.
You’re right, the community as seen on the home instance has all subscribers.
Is this a case of the servers just needing to sync with each other?
Thee is no sync code for remote user profiles and communities. The local server counts what it has, which is often an incomplete copy - and especially the users/subscriber numbers are not fully represented. Communities also do not automatically get discovered by a new server, a logged-in user must initiate a a search for a remote community for it to be known - and further at least 1 account has to subscribe for content to flow.
Just FYI, [email protected] (note the “-ed” instead of “-ing”) is far bigger than any of the communities shown in those search results.
That’s intentional. You only see subscribers in your local instance. You might be the only subscriber - if there are no subscribers then the community won’t be loaded in your instance, that doesn’t happen until a user subscribes.
The bigger issue is that every post and every comment is given a unique number based on the instance it’s in. It gets multiple numbers, one for every instance that’s federated with that comment. However there is no easy way of moving between comments and threads to the same on any other instance (except for the host instance). This makes it very hard to reply to a comment on someone else’s instance with your own accont in another instance - you have to log into your account, then find the comment on your instance. You can’t guess the link, you have to go to the community, find the post manually, then find the comment manually.
Posts and comments should just be assigned one number, the number of the user’s home instance. Then every other instance could just use that number with an instance tag, like with communities and users, eg
https://myinstance/post/1234567@theirinstanceYou can search for a post or comment using the full URL from the other instance.
But yeah, posts and coments should be assigned GUIDs. This was probably avoided because your local instance creates the post before it is federated the the home instance of the community. There are probably some technical challenges but I can’t imagine they are insurmountable.
As an example, you could say the instance with the person creating the post should assign the GUID, but then you could have bad actor instances assign already used GUIDs to intentionally cause things to break. You could say the communities home instance should assign it, but what ID does it get between when you post it and when it’s federated to the communities home instance (it doesn’t go straight away, but is queued for background federation - which isn’t always successful)?
These are things that will likely have solutions, but it’s easy to see how the current setup came to be. It’s mich easier to manage from a technical perspective.
