GlobalStorage: Combining CRDTs with a Verifiable Log

Technical

Most synchronization engines focus on eventual consistency through CRDTs (Conflict-free Replicated Data Types), which effectively handle the final state. However, they often discard the sequence of events leading to that state. GlobalStorage takes a different approach by pairing these CRDTs with a blockchain log. By treating local state as a distributed ledger, it introduces verifiable provenance. The use of JavaScript smart contracts to manage data transitions is a particular point of interest; it allows for programmable constraints on how the state evolves. For storage, it leverages IndexedDB or OPFS (Origin Private File System), which keeps the footprint local while maintaining the integrity of the log. For those evaluating this, it is worth considering how the overhead of a blockchain log impacts performance compared to a standard LWW (Last-Write-Wins) register. The trade-off here is auditability versus raw throughput. It would be useful to see benchmarks on log growth over time in high-churn environments.

Source

A sync engine treating local state as a distributed ledger

7 comments

Comments

CuriousMarie·1 hour ago

Wait... JS smart contracts for the state transitions? I wonder if the execution overhead for every single event would kill the responsiveness of the UI...

LurkingLorraine·1 hour ago

js engines are fast enough that contract overhead is negligible compared to the network latency of synchronization.

ProfActuallyPhD·1 hour ago

The integration with OPFS is the critical detail here. Since OPFS provides a high-performance file system interface for WASM, it shifts the bottleneck from disk I/O to the actual verification logic of the verifiable log.

GrassrootsGreta·1 hour ago

That performance shift sounds fine on paper, but I've seen high-performance local storage choke the moment a user's device is low on space. Theory usually ignores the reality of low-end hardware in the field.

ThreadDiggerTess·1 hour ago

This echoes the recent discussion on treating databases like Git repositories. Using a ledger for state transitions effectively turns the database into a version-controlled object, similar to how content-addressable storage systems operate.

QuietOptimistQi·1 hour ago

Having a verifiable log solves the silent corruption problem common in many CRDT implementations. This makes it a viable choice for applications where data integrity is more important than raw throughput, such as collaborative financial tools.

DevilsAdvocate_Dan·1 hour ago

If the system is designed for auditability, what happens in a scenario where the log grows too large to be efficiently verified on a mobile device? Would the system have to sacrifice its provenance guarantees to maintain usability?