"Christine what do you think of people running Bluesky relays now"

It's good. It's not decentralized, it's still a star topology (there's still one "main" relay everything consumes from (EDIT: I'm wrong, see followup)) and most of these are cheap by being shallow. But it's def good for people to do it

EDIT: The thing that is good is that people are running the infrastructure. I said above it's still not decentralized, and it *is* still decentralization-washing; that's the main risk of it.

@cwebber So…. Where is the good part?

I mean, if it is still all centralized, and they act as gatekeeper, I’m not sure where the good part of this is.

@Linux Good question. It's good that other people are trying out the infrastructure, because currently in theory nobody else can run the code (this is getting closer to people being able to do more of that). However, it's not good in that it still decentralization-washes things.

@cwebber I’m sorry, I’m still not seeing the good part. If it is centralized and, in theory, nobody else can run the code…

I think I like ActivityPub and how the Fediverse works. Anyone can use it, no one can gate keep it, and it is not centralized.

@Linux People are running code that performs part of the protocol. But it's still a problem.

At any rate, I updated the original post to clarify. As I said even in my original post, it's still not decentralized.

@cwebber In what way are fediverse relays re-centralization? They're not really comparable to Bluesky relays in that they're completely optional and there's no single central relay, they're just for smaller instances to get more posts federated to them.

@noisytoot It's not the worst. I am not *that* worried about fediverse relays. But I am worried about them being used as a solution for archiving, if used wrong.

More later. But it's not a big concern/threat. It's more that there's a better way I think.

@cwebber when you say that "in theory nobody else can run the code", which code part are you referring to exactly?

@laurenshof It was also awkwardly phrased: more that the code that was out there *wasn't* being usably run by the community, so this is iterating towards a point where more of it is, and more of the network can be operated by the community.

(Though my understanding is also that the private mega-nodes of Bluesky are also running a lot of code that isn't publicly available at this time, but I haven't looked into that in detail, would be interested to hear if that's right or wrong.)

I was wrong in my post, apparently the new relay demos are connecting directly to PDS'es.

If enough were connecting to the whole network, I think this would lead to the quadratic scaling issue I outlined in my previous blogposts. But anyway, I got it wrong.

It's a star topology in practice insofar as people are not largely relying on this, but it seems I was wrong to say that the new relays are a star topology. Mea culpa.