Whenever the AT protocol of BlueSky is set for others to deploy, won't it have similar problems as ActivityPub instances?

fediverse_report, 3 months ago

I actually set out to answer this question in a blog post, but it turns out that the answer is quite complicated, so I have to write an entire series about it. First part I published this week, which explains all the different components that make up the Bluesky network:

fediversereport.com/how-bluesky-works-the-network…

I don’t think that they’ll run into the exact same problems that AP-fedi has, as the design decisions are often made specifically to avoid some of these. However, their design decisions create new sets of problems for the network, which I’ll get into later

otter, 3 months ago

A friend was asking about reading materials for the fediverse, specifically around how different platforms are handling instance selection. I’m taking a look through your past blog posts, but would you have any links handy for me to pass along?

fediverse_report, 3 months ago

Oh thats an interesting question! I’m assuming you are talking about the UX/UI of instance selection?

And thats not something I have written about (neither does another article pop into my mind either sadly), but interesting idea for an article for sure to write about

otter, 3 months ago

Yes that’s it, and no worries! Looking forward to the future articles

maegul, 3 months ago

Nice! Was wondering if you’d pop in to this thread!

Looking forward to the rest of the blog series. Are you at all interested in or looking at what layers/components are amenable to 3rd-party community driven (decentralised?) development of alternative/new apps etc. The feeds system is obviously designed for that (and Relays obviously for big hitters only) but I’m unclear on the rest and curious to hear what anyone else thinks.

fediverse_report, 3 months ago

Heya! Good answers earlier by you!

Yeah I think I’ll have to get into that, but I’m starting to run into the limit of not being a programmer myself, and information is pretty scarce on ATproto. The article differs from their own federation architecture description from earlier in the year, simply because its outdated and noone has formally written down the new info, so that was a bit of a struggle haha blueskyweb.xyz/…/5-5-2023-federation-architecture

At any rate, the PDS’s are amenable for sure. Robin Berjon is the furthest along with thinking here, with his AP over AT piece: berjon.com/ap-at/ Responses I’ve seen havent suggested its technically impossible, but probably difficult for reasons that I tuned out of reading because I didnt understand :D

Beyond that, people keep talking about the lexicon and how that at is core is also versatile; similar to how fedi has Mastodon’s type=Note that everyone uses, even though you can create any ‘type’ you want. I’m pretty sure that nobody has done that yet tho.

sandwichfiend, 3 months ago

@ALostInquirer @fediverse The only thing bluesky has going for it are the domain aliases otherwise it's like diet twitter

Nighed, 3 months ago

What is the likelihood that you could set up a server to bridge the two protocols?

gashead76, 3 months ago

From what I’ve read of this, it’s basically already done. The developer is just waiting for AT Protocol federation. I’m not sure how to feel about how it works though, it’s got some quirks because of the differences in the protocols.

github.com/snarfed/bridgy-fed

spiritedpause, 3 months ago

I’m working on a client/app called Agora that integrates bridges like bridgy-fed so that you don’t have to think about those quirks, you just search something like “aoc.bsky.social” on it while logged in to a Mastodon account, it’ll automatically pull up the bridged version of the account for you to follow.

agorasocial.app

gashead76, 3 months ago

Oh wow, that’s really cool! I’ll definitely take it for a spin. I’ve been thinking that something like this would be great.

maegul, 3 months ago (edited 3 months ago)

Not an expert, I only have impressions …

The main difference is their architecture which is to some extent presumed by their protocol or at least designed to work with it. The architecture is more pyramidal while the fediverse is more flat/spherical. That is, ATProto will have big centralised servers receiving and sending out basically all the data, with everything else like the platform (like bsky, mastodon, lemmy) and feeds (like communities) and moderators (kinda like communities too?) processing and filtering this data stream for as it sees fit. Across these platforms, feeds etc, you’re able to move around with a single account and pick and shoes what you use. As you move further down stream from the big central server, the app you make becomes more simple to make (like the feeds in BlueSky now).

The idea being that individual instances aren’t so central to the system, which I think is a good thing. Users are more free in their system to move around without new accounts and losing or breaking the connections they’ve made. Also makes sense to me.

The big alarming aspect is the primacy of the big central server. They’ve said that multiple can exist but that it’ll be like search engines, only a few will bother to put in the effort.

By comparison, the fediverse is flat … a bunch of independent instances communicating with each other as they see fit. That’s basically it. Each user is bound to and stuck on their instance, which means when defederation happens it can be quite dramatic for the simple and obvious reason that all the users of an instance aren’t the same or hold the same values.

Frankly, it’s a system designed to create drama, while, IMO, mostly failing at its purported purpose, which is to foster communities. Because, first, you need software platforms to facilitate, and most on the fediverse suck at it (where are instances specific private chat rooms of any sort on the fediverse), and second, communities grow organically which is actually impeded by being stuck on an instance.

The latter is why we have federation in the first place … organic community building … which apart from basic moderation needs really calls into question whether the whole instances thing has much value apart from being the obvious Web 2.0 circa 2012 way of doing decentralisation. Taking the tight coupling between user, instance, platform, moderation (and allegedly community) but instead allowing these to be modular and composable with the user in the driving seat all sound like good ideas to me.

I personally hope ATProto mostly delivers on the promises of their system so that this whole idea of a decentralised social internet can move forward.

ada, 3 months ago

Bluesky isn’t decentralized though if your description above is correct

maegul, 3 months ago (edited 3 months ago)

So, again, I’ve got an impressionistic understanding.

The big central server, AFAIU, can be decentralised. How multiple such servers share data with each other, I don’t know. But, AFAIU, policy choices can be made at the level of the big central servers about what content or users they allow etc. But the idea is that moderation decisions will be made by downstream services. And the expectation of course is that the cost of running such a server would be prohibitively onerous except for a few instances. So it’d be a soft form, oligopoly like, of decentralisation you might say, if anyone actually makes one (currently there’s only the BlueSky big central server). It’s an interesting question though how small a community would be necessary to collect the wherewithal to start a new one.

Otherwise, the idea, AFAICT, is that it’s just a service and not intended to affect content and users much, which is instead up to all the downstream services which are intended to make up the actual user experience. The idea is that these layers will be open and decentralised. AFAIU, BlueSky is technically already decentralised now, under the hood, as to test things they’ve split the platform over multiple instances. I’m pretty sure their equivalent of an instance is something anyone can run (where BlueSky the app is open source). So it’s some form of decentralisation I suppose. I haven’t seen any breakdown though as to what community building facilities, tools and powers are available from the system. That’d be interesting, especially in comparison to the fediverse. I’ve seen it predicted that as BlueSky opens up there’ll be a migration of some sort to the fediverse as vulnerable users begin to suffer from abuse that the fediverse offers more protection from. But BlueSky is also a moving target.

I suspect the big eagle eye perspective is that the fediverse leans more “balkanised and independent servers” and BlueSky (or at least the promise/idea of bluesky) leans more “distributed public square”. In the push away from mega social media platforms, I think there’s probably a need for both to exist, at least in the interim as this whole “transition”, which is likely more lengthy and chaotic than many would like, pans out.

In the end though, bluesky feels like a more fragile project. They’re a bit “late to the game”, clearly have only a small team working on it, without a particularly compelling platform, and a little like the fediverse, have a slowly/slightly slipping away user base at the moment. There’s a real chance we won’t be talking about it in 2 years time. It could also do really well and suck the oxygen out of the fediverse in the same time TBH. In reality … this whole thing may just take a while to shake out.

RobotToaster, 3 months ago

I’m personally not convinced bluesky will ever be truly decentralised.

I suspect the reason they have delayed implementing federation for so long is to get as many users on their centralised service, and the big central server will be used to impose their choice of censorship. Even if you can theoretically run your own master server, if 99.9% of users only use bluesky’s it’s meaningless.

I do agree that defederation is the biggest issue holding the fediverse model back, however. It makes choosing a server overly complex.

maegul, 3 months ago

Not a bad take on BlueSky. The other funny side of it is that many of the users there don’t want any federation to happen at all.

It’s not unreasonable though that they have wanted to make sure their system of federation etc works well before they open it up, and that that has taken some time. By all accounts they have a small team and are generally making slow progress, and if they see themselves as competing with Twitter, they probably think it needs to work well straight out of the gate.

dumpsterlid, 3 months ago (edited 3 months ago)

It’s not unreasonable though that they have wanted to make sure their system of federation etc works well before they open it up, and that that has taken some time.

Yeah it must take a ton of work to make function properly but at the same time, I also think it isn’t unreasonable to doubt that a for-profit entity is ever going to willingly open up their system.

Why wouldn’t they just keep coming up with excuses for why it hasn’t been implemented yet? There is no real genuine promise here like there would be if the federation system was implemented from the start (even if it was janky) and every day that goes by the likelihood that federation/decentralization will ever happen on Bluesky becomes less and less likely.

This reminds me of an early access game selling itself on a vision of what the game will become while clearly having very little intention of ever getting there. They spend a lot of time talking about all the great features they will add like NPCs and quests but are they actually ever going to tackle the tough problems of implementing those features? Or are they just going to focus on selling character and weapon skins through loot boxes?

maegul, 3 months ago

All good points.

mozz, 3 months ago

I believe you've arrived at the crux of the issue

half_built_pyramids, 3 months ago

It’s almost like federation and everything that comes with it – plusses and minuses – is the point.

mozz, 3 months ago

At this point BlueSky had better make free liquor and hot BBQ come out of your computer if they want people to adopt it. I could be wrong, it could be real attractive, but I feel like there's so much momentum built around ActivityPub that they're by now fated to abandon AT for ActivityPub whether they want to or not, or else go the way of CompuServe.