The Great Decentralization Debate: Bluesky vs. Mastodon

I've been meaning to write a bit about this for a while, but finally getting into Bluesky last week has given me the push I needed to get to it. Back in November, Christine Lemmer-Webber, one of the original authors of the ActivityPub protocol behind software like Mastodon, wrote an excellently-researched piece called How Decentralized Is Bluesky Really? A lot of the technical ins-and-outs of the essay go a little over my head, but I think it's a worthwhile read if you, like me, are enough of a nerd to want to understand what's happening technically under the hood of new alternatives to privately-held social media applications. If you don't, this is probably the post for you, since this is my best attempt at reducing some high-level technical and design debate into something my dumb brain can comprehend.

The key takeaways here for me were that decentralization, to Lemmer-Webber, means anyone can feasibly host a node in a network, and that no node holds a higher tier than any other. The federation model allows for this through some neat technical tricks that, while making it hard to adjust to for Twitter users, mean in simple terms that it's significantly easier and less expensive to host than the firehose of every user in the network that services like Twitter and now bluesky provide. The flipside of this is that federated networks don't register much of what's happening across the broader "fediverse": you have to opt-in to federate with a server, and even then there are various technical tricks largely designed for privacy and memory purposes that mean your network doesn't has to host a copy of everything someone on another network has posted. This is great for meeting the goals of decentralization as defined, but not so great for discoverability or having a "global town square". Mastodon isn't very searchable, it's a bit clunky to use, and you're much less likely to do numbers on there than on a centralized service. It's more like a series of loosely-connected villages than a town square.

ATProto, the protocol behind Bluesky, offers the value proposition that it functionally replicates the global town square experience while being decentralized in the ways that they think really matter: allowing for "credible exit" and treating the company as a future adversary. Credible exit means that, even if a central node goes down, the information on the network can still be accessed. For "company as future adversary": Bluesky provides a global firehose that, while likely too expensive to self-host a node of, is built in such a way that any adequately-funded competitor could do so if, for example, Elon Musk buys it. I'm enjoying my time on Bluesky - it scratches the old endorphin receptors in a way Mastodon never has - but it wants to do a fundamentally different thing from Mastodon. It's mostly designed to create a version of Twitter that can't be entirely captured by one actor and isn't subject to the whims of a single server. If ActivityPub is anarchist, ATProto is anarcho-syndicalist: a hierarchical structure, but one designed to prevent single-actor control.

The interesting wrinkle here that I wasn't aware of is that, according to Lemmer-Webber, the way ATProto is built means that by its nature, the cost of running the network increases by adding more nodes. This creates a natural disincentive for multiple companies to run ATProto at once: the idea that ATProto could be a meaningfully decentralized network is highly unlikely in practice due to the cost, and while it still allows for folks to migrate to a node hosted by separate companies, there's still a high risk that everyone hosting a node is a malicious actor. Replacing a monarchy with an oligarchy may not be a significant improvement.

Again, a lot of the technical details here go a bit over my head, so apologies if I've gotten something wrong. The core takeaway is that, as things stand, Bluesky doesn't meet a strict definition of decentralization. I think credible exit and company as future adversary are worthy goals, too, and despite myself, it's nice to be able to see all my favorite sports and joke posters again. Mastodon has a long way to go before it can get a userbase that isn't primarily a specific type of nerd. But the risks Lemmer-Webber points out are legitimate concerns, and I think folks should be aware that, while I think Bluesky is trying to do something good, it may well fail. There just isn't a solution out there that's truly decentralized and not clunky and awkward to use at the same time, and it's possible those two goals are just incompatible.

Bryan Newbold, one of the lead protocol engineers at Bluesky, posted a thoughtful response, and Lemmer-Webber said what she thinks will be her last words on the topic in reply to him. Generally Lemmer-Webber seems to be correct to me, but again, I'm not a protocol engineer. This was a beautifully civil exchange between two people who clearly want to build something different, and again, both are worth reading if you want to get into the technical weeds. I hope the best for both of these projects, and I'm glad people much smarter than me are trying for solutions. In the meantime I will continue to be wherever dril is.