Actor Protection Commitment (APC) protocol [edit, we're renaming everything to "modular" due to our extreme emphasis on composability and extensibility. APC = Modular Web], is an API-compatible approximation of a later ideal protocol which mako is presently developing. The ideal protocol will be decentralized, but it will also be complicated and may take a few years to emerge, and mako wants to explore/demonstrate UX right now, so a much simpler centralized placeholder called APC will be used in the meantime. If the time comes to register an organization, we will put in place legal commitments to switch away from the centralized placeholder protocol by a certain time.
Why not federated? APC and its apps are open source, and the data is all self-verifying and either public or user-managed, so if we ever fail to host it responsibly, anyone else will be able to take over. We also think that once the complete protocol emerges, the concept of federation will stop making sense in general. Federation generally doesn't line up very well with the data caching needs.
But the main reason is just that centralized hosting is simpler to implement for now. Federation may be added later, given that we're already coding for full decentralization, it might not take much extra work. Idk. It doesn't matter yet.
APC
Features:
- Content-addressed distributed type system for defining schemas/API/mesa-standards in a resilient yet permissionless way. And it's a strong type system.
- (I think it'd be fair to say modular types is already a decentralized thing. The way data is hosted or transmitted isn't really part of the spec. You can do it over distributed storage. MW might not, but that has nothing to do with MT.)
- Ability to express cyclic structure within content-addressed data. You might not think you need it, but we needed it for the type system, and for some very basic permissioning datatypes, so we're guessing a lot of people are going to need it. I've used programming languages that are bad at/strictly forbid cycles and it never felt like a particularly natural restriction. Cycles occur often in nature. It's good if you can represent them.
- First class support from the neschat browser.
The ideal protocol of the future (is not APC as of today)
Today's decentralized protocols have been unable to circle the triangle of Fast, Reliable, and Cheap. Holochain and Freenet are Cheap and Reliable but not Fast; activitypub and atproto are Fast and Cheap but they lack verifiable finality of mutation; distributed ledger protocols like Convex or Solana are Fast and Reliable but not Cheap (or, they're cheap until they see mainstream adoption, at which point they will certainly no longer be cheap and might also no longer be fast). We believe that it's very likely that this tradeoff will be broken by ZKVMs by encoding the peering logic and network routing optimization into the core contract.
As soon as we can create or identify a protocol that circles the triangle in a way that can be abstracted well and made ergonomic, we will bless it with ports of our software.