Skip to content

spec.md

Latest commit

 

History

History
44 lines (31 loc) · 2.12 KB

spec.md

File metadata and controls

44 lines (31 loc) · 2.12 KB
Raw

object-storage Specification

Purpose

Lets the backend store and read objects through the S3 API alone. Moving from local S3-compatible storage to S3 in production is then a configuration change, and storage never accepts unauthenticated access.

Requirements

Requirement: S3 access configured only by properties

The backend SHALL reach object storage only through the S3 API. The endpoint, region, access key, secret key, path-style setting, and bucket name SHALL come from configuration properties, not from code.

Scenario: Store and read back an object

  • WHEN the backend is configured with a storage endpoint, valid credentials, path-style access, and an existing bucket
  • AND it stores an object under a key
  • THEN reading that key through the same client returns exactly the stored bytes

Requirement: Storage rejects unauthenticated access

Object storage SHALL reject requests that carry no credentials or wrong credentials.

Scenario: Anonymous read is rejected

  • WHEN a client with no credentials requests an object from the bucket
  • THEN storage responds 403

Scenario: Wrong secret key is rejected

  • WHEN a client signs a request with a valid access key and the wrong secret key
  • THEN storage responds 403

Requirement: Bucket exists before the backend starts

The local stack SHALL create the configured bucket if it is missing, and the backend SHALL NOT start until that step has succeeded. Running the step again SHALL succeed without changing anything.

Scenario: Bucket ready at startup

  • WHEN the stack starts from empty storage
  • THEN the bucket-creation step exits 0 before the backend starts

Scenario: Repeat run is harmless

  • WHEN the bucket-creation step runs again against storage where the bucket already exists
  • THEN it exits 0

Requirement: Storage creates no buckets implicitly

Object storage SHALL NOT create a bucket as a side effect of writing an object.

Scenario: Write to a missing bucket

  • WHEN a client writes an object to a bucket name that does not exist
  • THEN storage rejects the write instead of creating the bucket