Hm? Which part of the stack requires a lot of resources? Each part (PDS, Relay, AppServer (AppView)) can be run on a PI. Blacksky (not a large organization by any means and completely community funded) is almost running the entire bluesky stack themselves. IMO there's not much of a reason to run every part of the stack yourself unless you want to be completely independent from bluesky. At the very least if you want to host your own data then that just means running a PDS, which means finding the cheapest VPS you can that fulfills the min spec. Mine costs about 7 bucks per month and I never have to think about it. I'm also running a relay for no reason other than for learning purposes for about 30 bucks a month.

The AppServer, if you want index the full bluesky network (39M users) will run you about 200-300 dollars a month. Again, not really needed but you can if you want. There's also experiments people are trying out to only index smaller parts of the network, e.g. only users you follow which would mean hosting an AppServer would be even cheaper. FWIW, I like both protocols and want them to succeed, anything that gets people off of closed social media.

It sounds like the AT protocol is improving in the right direction, but the numbers still say it's highly centralized and therefore closed social media owned by one monolithic entity.

AP has its issues, namely not being at all consistent, but that trade-off allows anyone to run all the components of the network without breaking their bank.

AP is just “many small centralized services that email each other”. Its scaling characteristics reflect that - of course running a “little Twitter for 100 people” is cheap. But it’s a completely different thing. You could “scale down” atproto in the same way, but the point of atproto is that you can aim higher than niche islands.

ActivityPub doesn’t attempt to solve any of the same issues that atproto does — there’s no ability to have a full consistent view of the network. So it’s comparing apples and oranges.

I agree that AT aims high, maybe a little too high.

To me Fediverse is basically all the forums we had in the 90s and 00s, but now they can talk to each other. So with that said, I am principally against huge instances like mastodon.social and such.

While bsky is more like decentralized Twitter, meaning it also requires a significant chunk of the resources of Twitter to run consistently. Which is also why it has not decentralized yet, and probably won't any time soon.

We have to ask ourselves, what is the point of decentralization? What is the USP?

To me the USP is that no one person or entity can buy it, sell it, or ruin it.

If we had to start over from scratch then I think AT would be a good way to start, because it would presumably not be federated with bsky and therefore low volume. But as volume grows it becomes harder and harder to maintain for small groups and eventually they all consolidate into big groups that are easier to take over and ruin.

Basically it's a difficult decision we have to make, a trade-off between consistency, and decentralization. Which do you value more?

Me never having been on Twitter, and coming from that era of the 90s and 00s, I value decentralization and small groups participating in a federated network more than large entities.

How I envision the future of social media, all these actors like bsky, Twitter, Meta, Fediverse will continue existing side by side, but Fediverse will likely be the smallest and most niche of them all. Fedi has, to me, taken the place of all those old message boards I used to hang on, while the rest are mainstream social media that made its entrance onto the world wide web back in early 00s with Facebook.

Depending on what exactly you mean, this isn't the case. For example, running your own PDS is very cheap.

If you want to fully run a full copy of everything yourself, it's going to be more expensive, sure, but those costs have gone down dramatically over time. The most expensive bit is running $34/month: https://whtwnd.com/bnewbold.net/3lo7a2a4qxg2l