TigerFS and PostgreSQL: correcting a TigerBeetle mix-up

Correction, September 16, 2026. This article mixed up two projects. We have withdrawn the claims about a supposed deterministic TigerBeetle storage layer and the recommendations based on that confusion.

Which project we are evaluating

TigerFS, in Timescale's repository, provides filesystem access to PostgreSQL. Its README demonstrates tables, rows and columns accessed through paths and tools such as ls and cat. That is the proposal to evaluate here.

The useful question

My evaluation criterion would be specific: does this simplify a task that currently requires a SQL client or a custom integration? A project's name is not enough to infer its guarantees or architecture.

Before adopting it, I would propose a bounded test:

1. Choose a read task against a test database without sensitive data.
2. Record the project version and database schema.
3. Compare filesystem results with an explicit SQL query.
4. Review permissions and write semantics before allowing modifications.
5. Document failure behavior and the additional knowledge maintainers need.

That test has not been executed for this publication. This article contains no performance results, measured recovery guarantees or personal production experience.

What changed from the previous text

The recommendation based on financial accounting and byte-for-byte reproducibility has been withdrawn: it did not follow from the correct source. The reasonable next step is to validate one concrete task with TigerFS, recording versions and results, before recommending adoption.