Bluesky made some key decisions that are different from Mastodon:

- Profiles and posts are portable across hosted instances. If you have a problem with your admins, etc, you'll be able to hop away without much fuss.

- Moderation is decoupled from hosting. They're still working out the details, but the overall vision is that the global firehose is public, and you'll be able to opt in and out of different moderators which operate on that stream. If you don't like what a moderator is doing, you'll be able to just turn them off for yourself. Moderation doesn't prevent people from posting, it allows people to prevent themselves from seeing.

Which is all good in principle, and it'll be interesting to see how it works out.

The profile portability story is good, save that as far as I can tell, Bluesky server operators don't actually do much save for provide storage and a place for clients to connect to, specifically because the algorithms and curation are being treated as a separate layer. So ironically I see far less reason for identity portability in that ecosystem since (to my knowledge) the server you're using doesn't have much of an impact on your experience.

Additionally, this layered functional architecture creates new kinds of problems.

For example, with Mastodon, if I'm helping fund my instance (which I do) I'm also helping fund the moderation staff, because the instance architecture and the moderation story are tightly coupled.

With BS, moderation is a separate service which has to be funded... how?

Same goes with curation.

So I still have a lot of questions about the Bluesky model.

And this is all assuming they actually pull off what they claim, and it's still waaaay too early to say if this'll all work in practice.

> So ironically I see far less reason for identity portability in that ecosystem since (to my knowledge) the server you're using doesn't have much of an impact on your experience.

One of the biggest reasons why people change Mastodon instances is because the instance shuts down.

I had to do an emergency switch because my instance admin went AWOL, and it was breaking down technically. In other cases, the admins decide they don't want do run an instance anymore, for whatever reason.

If posts could be migrated on Bluesky, that would be a vast improvement over Mastodon.

You are right, that is definitely one very good reason why many people move (though in the Mastodon world, I would argue it's probably not the biggest reason outside of weird events like the shutdown of mastodon.lol), and in the Bluesky world, probably the primary reason (since there's no concept of local moderation or community policies that you might disagree with, thereby leading to an instance switch).

> I would argue it's probably not the biggest reason outside of weird events like the shutdown of mastodon.lol

A number of such events have occurred.

I haven't actually heard of anyone moderated off an instance (aside from spammers), but maybe I hang with the "wrong" crowd for that.

It absolutely happens.

Eh, tbh though, I don't know why I was arguing that point, it really doesn't matter one way or another.

You absolutely identified a very good reason to want to switch instances, and why account portability has value in the Bluesky world.

I wish ActivityPub had a better solve for this, but unfortunately a true federated model makes that awfully difficult (since the entire ecosystem has to update its view of where posts are located).

Similarly, if I'm in a length email exchange with a friend on Gmail, and then I move to another service, there's no way to update my friend's account to "rewrite" those emails to point to the new service. It's a deep limitation in the model itself. The only question is whether the tradeoff is worth it (and, to me, it is).

Another very cool decision they made for AT Protocol/Bluesky is making user identity domain name-based. So you can just sign up as a subdomain of eg .bsky.social, but you can also bring in a custom domain name. This neatly solves identity-portability and even trust/verification in one swoop

(you can also change which domain your bsky profile is identified by, which, I'm not quite sure how that works at a protocol level but it is possible)

> This neatly solves identity-portability and even trust/verification in one swoop

Until you lose control of the domain for any number of reasons and then whoever buys it happily continue on as you. A simple credit card processing problem can lose your "identity" at some point in the future.

The same is true if you use a custom domain for email, for your company's servers, for any number of other critical things. It's no different (and arguably those other things are more catastrophic than losing control of a social media profile)

Also, to emphasize, regular people who don't do those other things don't have to roll a custom domain on bsky (and are unlikely to). But eg. if NPR wants to make a bsky account under @npr.org, that's really quite good verification that it's "the real NPR account", and nobody has to deal with any of the politics around verification which have become a dumpster-fire on twitter since Musk took over

Using a domain for verification isn't the worst idea but there's still some things to work out. How will the service handle homoglyphs domains? Are they going to ship a webfont that makes homoglyphs obvious? Will they display punycode? It's all well and good the real NPR can use @npr.org but what will ｎｐｒ.org end up looking like on the service? Will it be distinct enough normal users will be able to distinguish it from the real domain as displayed on the website or app?

Again, it's a strict improvement over what came before (centrally-managed namespace), and it lumps bsky's problem in with the existing place where the web is already wrestling with these questions. It hitches its cart to the world's biggest and most robust decentralized-identity mechanism; I can't think of a better approach to take, perfect or not.

Actually the domain (handle) resolves to your DID which is you primary identifier and which contains your public key and without coresponding private key, nobody would be able to happily continue as you... See: https://atproto.com/guides/identity. It would be better though if the DID didn't contain the link back to the domain, so that you could continue with your identity under different domain...

> So ironically I see far less reason for identity portability in that ecosystem since (to my knowledge) the server you're using doesn't have much of an impact on your experience

To an extent, though I still think it's valuable for situations where eg. a host might decide they don't want to host you if you post certain content. In the same way: a website host doesn't normally have much of an impact on your experience, unless they decide to kick you off or otherwise exert control over you

> moderation as a separate service, which has to be funded... how?

It's an interesting question; I think some of it will be answered with algorithmic/AI-based filtering (false-positives are more tolerable if users have the ability to opt out!). People have also talked about a model where you can just share (or subscribe to) communal block-lists. So maybe your circle of friends does its own internal "moderation" just by using the site (blocking people is already a normal part of my day today usage of both bsky and twitter)

With a platform this open and flexible, there are tons of options to be explored!

> And this is all assuming they actually pull off what they claim, and it's still waaaay too early to say if this'll all work in practice.

It definitely is! But I'm excited and hopeful

> If you don't like what a moderator is doing, you'll be able to just turn them off for yourself. Moderation doesn't prevent people from posting, it allows people to prevent themselves from seeing.

This is going to polarize people even further.

Before social media, had you seen people with radical views? Yes there were cults but it was all limited. Social media polarizes people and with this kind of moderation it seems it's only going to exacerbate the problem.

There will be no one left in these polarized groups to correct their course.

Edit: It's going to create more and more groups who don't want to connect and have a discord on common issues.

a) You could say this about any communication channel- the web, email lists, in-person meetings. Should we require a moderator in every social sphere?

b) Personally I think butting heads and being outraged by what others are saying is the bigger factor in polarization, which this would actually help with (at least, I've noticed I became a lot more politically nuanced and a lot less emotional when I made a concerted effort to pare down my twitter feed using blocks and mutes)

How is this any different than servers on the Activity Pub network?

> There will be no one left in these polarized groups to correct their course.

This presumes polarized groups start neutral and can somehow being corrected away from polarization. People often seek out polarized spaces and then get more polarized but they're starting far from a neutral position.

>Profiles and posts are portable across hosted instances. If you have a problem with your admins, etc, you'll be able to hop away without much fuss.

As far as I can tell, this is only with cooperation of the server owner; since their "DID Placeholder" relies on a server to provide a string of identity updates. ActivityPub can do the same thing with HTTP redirects; and Mastodon supports cooperative profile redirection, which is morally equivalent. Polycentric[0] is a tad more decentralized than this, AFAIK it's specifically designed to make cryptographic keys first-class identities (so called "sovereign identity"). No clue if they've managed to solve the obvious pitfalls of cryptoidentity.

[0] Another censorship-resistant social networking tool being funded by FUTO Tech, of whom Louis Rossman is a spokesperson for

Handles are DNS names that point to the DID

https://atproto.com/guides/identity

That moderator idea seems huge to me. That means Moderation is less like a benevolent dictator that people might disagree with, and more like a view filter.

There can then be many ways to implement the filters: voting, dictators, AI, etc.

Yep! Bluesky's solution to many of Twitters problems and controversies is "give people the tools to solve them how they want to for themselves"

It seems unworkable to me. If you filter the trolls, but the people you follow don't, your feed isn't going to make sense. This causes a race to the bottom in terms of moderation, which causes a bad experience for everybody.

One thing that moderation should also do is restrict which accounts can see your posts in their discovery feed. Even if the posts are public and those account would be able to navigate to the posts with a link sometime some of your followers are just there to spam/troll.

> At this point they're just kicking the can down the road, postponing the problems that Mastodon already had to face.

Seems they're finally publishing more details about it: https://blueskyweb.xyz/blog/5-5-2023-federation-architecture

> The federation architecture allows anyone to host a BGS, though it’s a fairly resource-demanding service. In all likelihood, there may be a few large full-network providers, and then a long tail of partial-network providers.

The blog post seems kind of hand-wavy. One question is, how do they make the few large BGS censorship-resistant?

(I know I'm violating the HN guidelines, but I hate how they've hidden the scrollbar on that blog post.)

Maybe its because of my crypto years, but nothing tunes me out of the conversation more than "decentralized".

There is no such thing. You can try your best but the laws of human behavior pretty much ensure that centralization and cartelization will happen, unless specifically stopped via outside regulators.

The best you can get is something like DNS or email, which are "decentralized" in theory but quasi-centralized in practice.

Which is much better than other systems; "twitter but like email" is a reasonable goal.

You can get more decentralization than those services provide, but not without sacrificing usability and/or security.

https://wikipedia.org/wiki/Zooko's_triangle

One key to a successful decentralized system is not being able to spam with it; if someone learned how to spam using BGP or DNS we'd all be screwed.

I think when many people say decentralised they mean federated, or something along those lines.

Email and IRC today are, as far as I know, highly centralised, but not in the same way that, say, social networks are.

Even the web is somewhat centralised (and I suppose always has been, to some degree) but it's clearly and successfully federated.

Mastodon's user experience issues should be perfectly resolvable - you can follow accounts on other instances, I don't see why you shouldn't be able to create accounts on other instances as well.

Go to an instance, select a bunch of interests, and have it return a set of instances that are accepting new accounts.

It's meant to be a federated service - it should act like one.

If mastodon.social was the entirety of mastodon it would be pretty bad and I wouldn't be using it. But oh well, guess people just want Twitter 2.0.