I thought about Svelte + Tauri. Desktop but with Web flexibility. It also allows adding other fediverse as plugins.

I read it and post from jebroa. At least latest alpha can do it.

Ok, you are right about peering, I tried to get more peers to be faster, but it isn't necessary. I didn't find anything about ActivityPub broadcasting, but if it's true... so, yeah, having rpc p2p connection doesn't make the whole system less federated. But still, usually crypto clients has lists of nodes (or api balancers) for faster handling.

Huh, so the problem is about just serving static assets? TBH, I think this problem should be mostly solved, especially for such minimalistic UI. Maybe some (free) CDN? Also UI can use any lemmy server for most of requests (e.g. fetched federated_servers fom a bootstrap node) and use "logged one" only for user actions. I think it isn't a terrible difficult task for the current ui (it has it's own backend).

I know what you mean, but all nodes are equal, they are fully participating (stay aside validators). I mean every every node handles every transaction and can be faster than another (it doesn't matter due to validation scheme, but technically speaking all nodes handling every user action)

Because we are here because of content, made by users. I'm thinking about whole "lemmy-verse". If users encounter issues, they just stop using the service. You as an instance owner can choose to not participate. But if somebody already thinks rhat they helps, why not use it?

I think a amall hit of p2p can be useful. Maybe as an addition layer. I worked a lot with tendermint nodes (cryptocurrency) and i saw pretty effective solutions.

Yeah, and this post about how to use some (a lot of) servers that are doing nothing to participate in "pros" while the top 20 of servers are suffering from these cons.

I dont suggest adding a centralization =) I see two possible and actionable directions:

1. Create tech solution to balance load through available resources
2. Spread the word that there are better ways to spend your money and passion helping lemmy. I know, my "engineering manager" bias tends to see process problems in places where are no problems. But I dont want to see how the awesome idea is dying because of lack of basic management and foreseeing.

I'd like to help with this improvement. Do you know any plans for it? Honestly, looks like that there is no "lemmy committee" and even lemmy's developers cannot organize something like this. Any ideas?

Yeah, it's nothing about communities. Technically speaking, only the amount of direct HTTP requests matters. If nobody opens your domain, your instance is just spending your money for nothing.

> Because you can still access all content no matter where you are.

If you know how and want to do it. Unfortunately, it isn't the way how most people think.

As I said, there is no profit from empty instances. Of course, the federation itself is good and fail-proof in this way. But if nobody asks for this cache, it's just an Internet Archive of a sort.

Yes, you are right. If this instance has members. A server will actively fetch "foreign" content and cache when this instance's user asks. But aside of top 10 servers, there is no profit of having more until they have a couple of dozens of users. If any server would have been able to "delegate" request handling to less busy servers, it will be a solution for this uneven load.

Please elaborate, how is "every instance is sharing in the traffic to browse the fediverse". I didn't find it nor in AP standards, nor in activitypub_federation lib docs. If there is some mechanisms of balancing inside the lemmy's code, would you mind pointing it for me?

There are two points:

1. Instance owners aren't ready to be a community manager and moderator. Also, a big instance costs much more. Many people just want to pay their 5$ in month and passively make their contribution.
2. Not all users are "power users". They want to consume content. To make federation successful, we should provide a smooth UX. Balancers, meta-account, something like this. Smart enough to require only one click and utilize all available resources.

Yeah, this is exactly my point. But I suppose a lot of people think that starting new instance is enough. But, unfortunately, it isn't.