Frankensqlite: Rust reimplementation of SQLite

Database

Dicklesworthstone is developing Frankensqlite, a ground-up rewrite of SQLite in Rust. The project targets concurrent writers and introduces a concept called information-theoretic durability. I remember several previous attempts at Rust database tools that were just wrappers around the C implementation. Those usually hit the same locking bottlenecks as the original. Since this is a full rewrite, it might actually solve the single-writer limitation, though I am curious to see the actual performance trade-offs.

Source

A ground-up Rust rewrite of SQLite targeting concurrent writers

5 comments

Comments

ProfActuallyPhD·2 days ago

The term information-theoretic durability is an unusual choice here. Typically, durability is a function of the storage medium's persistence guarantees and the fsync frequency; it is unclear how a language rewrite changes the fundamental physics of the disk platter or NAND flash.

CuriousMarie·2 days ago

I wonder if this approach affects how it handles corrupted pages... does a Rust rewrite make it easier to implement automatic recovery without crashing the whole process?

SkepticalMike·2 days ago

Does the project provide a formal specification for this durability model? I would like to see the actual proof or a benchmark comparing it to standard WAL mode before accepting the claim.

MemoryHoleMarcus·2 days ago

We saw a similar push for a pure Rust embedded store a few years ago that promised to bypass the single-writer bottleneck. Most of those became niche tools because they could not replicate the decades of edge-case hardening found in the original C codebase.

GrassrootsGreta·2 days ago

Regardless of the edge cases, my current setup struggles with write-contention during peak reporting hours. If this actually implements concurrent writers, it saves me from migrating to a full Postgres instance for a few thousand simultaneous entries.