9 posts filed under core.
Most database auth lives in middleware, which means every code path that skips the middleware skips the auth. MongrelDB stores Argon2id-hashed users and roles in the catalog and can enforce permissions on every transaction at the storage layer itself.
MongrelDB reconstructs change events from its own WAL and streams committed records to follower copies, so an embedded database can participate in replication and change data capture without bolting on a separate toolchain.
MongrelDB binds idempotency keys into the commit log itself, so a retried write is safe across process crashes and restarts, not just across network drops.
MongrelDB commits a transaction with a single fsync on the WAL, because the WAL is the only part of durable storage that has to be synchronous.
One Rust crate runs both the embedded library and the server daemon, and the seam between them is narrower than you would guess, which is the whole point.
SQLite SEE is the paid benchmark for database encryption at rest. Here is what building your own encryption layer costs in throughput and what the tradeoffs look like in production.
Learned-range indexes beat B-trees on cold sorted range scans up to 3-4x on p99, but only when the workload is recency-heavy; here is when the tradeoff is right and when the B-tree still wins.
MongrelDB's HNSW index lives in the same engine as the rows it covers, so vector search shares the same WAL, transaction, and /kit/query path as the rest of your data.
MongrelDB's full-text search uses an FM-index, not the inverted index you usually see, because the FM-index lives in the same WAL as the rows it covers.
