Comparison between Muse and Secure-Scuttlebutt

These are the last two event-driven database systems I, Henry, worked on. Contrasting them might give us some insights about the choices we want to make for Mass Market.

System Comparisson



Server holds all User Log’s

One Log per client/feed

Log is shared between all devices of a User

no multi-device => we want multi-clerk

Writes happen as RPC to the Server

writes always happen locally, which allows for true p2p between local clients

Local-first via “pending” writes

works b/c no complex shared state between feeds. Threads of posts are grow-only sets, in a sense.

Let’s consider the Muse model for us: The Relays would hold the write primitive, which mutates the Listing finally.

Stores/clients would fetch the mutations/events from a Relay and apply them locally, to reconstruct the state and verify against on-chain commitment data.


The Relays build the Log, Store users write to them.

The important distinction here is that the state of listing is the combination of all mutations applied, not the latest/longest chain.

Let’s assume we have two clerks working at a store: everything should be fine until they change the same item.

Mutating the listing