Hi Lemmy!
I'm studying webdevelopment, i love the concept of the fediverse and activitypub, but i still see it as really new and needing to get better of course.
I will just throw out some first impressions, these are just from a first sight so i'd be happy if you make me notice i'm saying something false or inaccurate.
As i said in the title i would really love to collaborate, so i'm just exposing stuff i would love to work to, just to see what others think about it and if they could be theoretically possible.
The Fediverse claims to be highly interconnected, yet i cannot use the account i have on a mastond instance on Lemmy, and neither the opposite. I don't see why there has to be many instances of the same thing and so many accounts needed to access, for example if i want to browse Kbin with my Lemmy account, i won't be able to, but i can create a new account and find some content of Lemmy on it or browse privately.
The system of mentions is quite confusing as well and i don't see these differentiation of istances a pro for the network, if an istance goes down the content on that istance goes down as well.
More that having to reference each server istance, it would maybe be an idea to have one single domain holding all the istances, much like IPFS and blockchain nodes. More servers hosting the content for the same platform, backed up by a pinning service such as Pinata for IPFS, not different servers holding different content loads.
I want to be able to subscribe to one single Fediverse account, and be able to use it in every different federated platforms.
I'll take the user login as an example: If i login in a federated platform, IPFS would perform a node lookup and find my account details on one of the federated servers, instead of having to specify which server i want the account info to be coming from.
All of this would still need to be backed up: 1- because node lookup is quite slow 2- because if one or more istances go down, there has to be a reliable and fast backup available to keep the platform running, even if the content is already duplicated in more IPFS servers.
Of course with this option there HAS to be a central server with all the data on it in order to make them readily available and make the user experience acceptable.
Another possibility would be using really light languages (such as Rust that doesn't have a runtime for example) to build the platforms and see what the performances are WITHOUT a pinning service and so without the need of a central server.
Please don't be harsh on me, i'm just jotting down some ideas!
Sorry i didn't explain myself very well.
By central server i meant a pinning service, but that could even be hosted by the already existent fediverse istances.
I'll paste one answer i added below:
I meant sharding. that’s what IPFS is about, the files on it are sharded between different computers accross all over the world that request the specific resource. When you ask through ipfs for a resource, it would perform a lookup and see if that’s available and in which node, then serve it to the client and cache it on its node. Content can be pinned to specific nodes in order not to be deleted through garbage collection, and that was what i was meaning by “central server”, there are services (like Pinata) doing just that: pinning content on fast servers and making it readily available. But i guess those nodes can also be the current fediverse istances, i had a central server in mind more due to that and because of the fact that smaller servers (available to the little associations that host fediverse istances) couldn’t probably keep up with something like sharding right now, but i don't really know.
I think something very similar is done already. If I look a thing up on another instance - my instance saves it. If the origin goes away - I can still see it.
Not the kind of sharding you're talking about, but the federated nature of stuff does prevent something popular from just dropping off the surface of earth.