What does sufficient decentralization mean?

I often see people say Farcaster isn't "sufficiently decentralized".

You're free to define what "sufficiently decentralized" means to you, of course, but how we think about it is laid out in Varun's blog post:

A social network achieves sufficient decentralization if two users can find each other and communicate, even if the rest of the network wants to prevent it. This implies that users can always reach their audience, which can only be true if developers can build many clients on the network. If only one client existed, it could stop users from communicating. Achieving this only requires three decentralized features: the ability to claim a unique username, post messages under that name, and read messages from any valid name.

You might ask: "But if most people use Warpcast, doesn't that mean it's de facto centralized?"

Depends on the perspective.

For a developer, if you are building on the protocol, then you're free to sign up your own users or convince users to move over to your app (it's easy for users to use multiple clients at the same time).

For an account with an audience, let's say you were unfairly nerfed by Warpcast, i.e. you don't appear on the app. There's definitely a lot of friction to get people to start using another client if they already don't, but the protocol itself is not restricting your ability to communicate. That's an important distinction because on web2 / centralized social networks, the app and the "protocol" (i.e. the database) is the same thing.

So if you're argument is Farcaster is not sufficiently "decentralized" (or "decentralized" if you don't like the sufficient framing) then you're defining decentralization as a protocol or network that has at least 2 independent, roughly equivalent functionality and UX clients (because majority of consumers revealed preference is UX over principles).

That's a reasonable definition.

But in practice, in the near-term, there's unlikely to be multiple, well-resourced enough clients that are full-time building on the protocol for amazing UX. That's not a knock on people building, just the reality that in 2024 consumers expect a lot of our their daily driver apps.

In the medium to long-term, dedicated individuals, small teams or open source projects can likely hit the UX bar. But that will probably take years of sustained effort.

So protocols in the early days are unlikely to hit the bar of 2 independent, roughly equivalent and functionality UX clients.

This is something I've changed my mind on over the years.

"But if Warpcast didn't have exclusivity on DCs and channels, then other clients could better compete"

Yes, but having both be interoperable / in the protocol doesn't magically mean all of that surface area gets built out in other clients.

I would assume other clients can move quicker by copying whatever patterns Warpcast has iterated to (avoiding all of our mistakes) and then in some cases, improve the UX.

But building out a robust chat UX in 2024 that's passable (not even Telegram level snappy) is a lot of work. That's effort not being spent on differentiating features or other core parts of the app. Same goes for channels.

And arguably, the most important part of a feed-based social app is the feed, which is mostly backend (machine learning, performance).

One advantage for other clients are dedicated infrastructure companies like @neynar, @openrank etc.

But it's still a lot of work.

To close: it's easy to have a bunch of opinions on how the ecosystem should be built. But if you talk to the people building the apps, they often have a different perspective having been in the salt mines.

That's not to say we are doing everything perfectly (we're not!) and we don't make mistakes. One of the reasons we are holding Developer Day next week is to get candid feedback from a group of developers working in the ecosystem daily. Goal is to get everyone better aligned and rowing in the same direction to make Farcaster successful.