How do "instances" and data replication work?

[trosel]

A public instance is available at https://rostra.me/.
Anyone can host their own public or private instance, or run Rostra locally.

If the default behavior is p2p, then how do instances affect that? If we use an instance, are we pinning our data to the instance database, so to speak? Is the source of truth of our own data still our devices?

Also, data replication is mentioned regarding p2p design. Are we replicating our friends' data on our devices as well?

[dpc]

If the default behavior is p2p, then how do instances affect that?

Instances are replicating data. Data forms a DAG and replicates, merges and diverges freely between instances towards unification.

Is the source of truth of our own data still our devices?

The source of truth are the cryptographically signed events. Instances only facilitate propagation, and compose it into usable view/UX.

Are we replicating our friends' data on our devices as well?

Yes. Replicating your WoT is a crucial part of improving reliability and availability. The assumption is that if you follow someone, they are worth replicating and helping to synchronized their data. Without it the system wouldn't really work assuming users's instances typically are offline.

[trosel]

Instances are replicating data

Can I use more than one instance at a time? Or it’s only meant to be one at a time because the client UI is combined with the replication?

The source of truth are the cryptographically signed events

And those live primarily in browser storage on your device? Meaning your device storage is the source of truth?

Replicating your WoT is a crucial part of improving reliability and availability

How much storage is allocated to replicating your friends? Is there any way to know who is taking the most space or how to better manage that?

[dpc]

Can I use more than one instance at a time?

Sure. It wouldn't be enlightened if you couldn't. I do it all the time, switching between local instance and rostra.me . They just syncs between each other like any other node, and thangs to a DAG data model there's no problem with ordering and merging diverging threads.

And those live primarily in browser storage on your device? Meaning your device storage is the source of truth?

Every instance replicates locally its WoT. Not sure if I would call it a "source of truth". It's just this instance's view on the state of the network. The truth is in the public keys and signatures, IMO.

How much storage is allocated to replicating your friends? Is there any way to know who is taking the most space or how to better manage that?

There currently isn't any limit. Each instance is tracking the stats per id, and architecturally it would be easy to add some limits and automatic prunning of old data, but since Rostra doesn't really get all that much public usage, I never got to it.

[trosel]

They just syncs between each other like any other node, and thangs to a DAG data model there's no problem with ordering and merging diverging threads.

What exactly initiates the syncing? When I log into a client on a different server, it kicks off a search and sync? How does it know which other server(s) to sync to? Does my pkarr address show which servers I use?

Not sure if I would call it a "source of truth". It's just this instance's view on the state of the network. The truth is in the public keys and signatures, IMO.

If the truth is in the public keys, then the public keys would be where you go to retrieve all the data associated with the public keys. But there’s no data stored in the public keys.

One device or server must have all the data, or at least more data than others, right?

I apologize if I’m not asking good questions. I know how SSB and Nostr work, but I’m not understanding rostra.

[dpc]

What exactly initiates the syncing? When I log into a client on a different server, it kicks off a search and sync? How does it know which other server(s) to sync to? Does my pkarr address show which servers I use?

Pkarr address is used for bootstrapping, yes.

One device or server must have all the data, or at least more data than others, right?

No one has to have anything. If you (or other people) have no instances keeping around your data, it is simply gone.