Declare reef-storage as substrate_kind object-storage. Hold non-secret attribute paths for Scaleway S3. Do not host a compute rail.
31 lines
1.1 KiB
Markdown
31 lines
1.1 KiB
Markdown
# INTENT
|
|
|
|
## Why This Repo Exists
|
|
|
|
`reef-storage` is the substrate boundary for **provider-delegated
|
|
S3-compatible object storage**. The first instance is Scaleway, used for
|
|
postgres backup and other large, slow-tolerant volumes.
|
|
|
|
We consume the S3 capability. We do **not** host a rail and we do **not**
|
|
operate how Scaleway provides or runs object storage. This reef names the
|
|
boundary, holds the committed non-secret attributes, and is the place
|
|
operators look for “where is our object store?”
|
|
|
|
It is a different reef from `reef-railiance`. Backup is not bolted onto
|
|
the home compute substrate.
|
|
|
|
## What This Repo Must Protect
|
|
|
|
- the split: procure/cost in `resource-control`, operate-by-reference here,
|
|
secrets in `railiance-platform`, consume in `rapp-*`
|
|
- no rail invented to satisfy a compute-centric schema
|
|
- no secret material in git
|
|
- independent failure domain from Host Europe / `reef-railiance`
|
|
|
|
## What This Repo Is Not
|
|
|
|
- a Scaleway account manager or billing ledger
|
|
- a Barman/CNPG operator (`rapp-postgres`)
|
|
- a secret store
|
|
- a generic S3 client library
|
|
- an extension of `reef-railiance`
|