Sovereignty is settled at the bottom of the stack.
All capabilities · Running
Sovereignty claims are settled at the bottom of the stack. A community that cannot run the system, patch it, observe it, back it up, and move it is depending on someone else's goodwill regardless of what the license says.
the products a community runs
identity, policy, secrets, observability
block, file and object, with stated durability
addressing, boundaries, ingress
general and accelerated, pooled and scheduled
sites, racks, machines a community can hold
What the substrate provides
Compute and scheduling
General and accelerated compute, pooled and scheduled, so workloads land where they fit rather than where someone remembered to put them.
Storage with stated durability
Block, file, and object storage with explicit replication, snapshot, and retention behavior, because "it is stored" is not a property until you can say how it survives.
Network and segmentation
Addressing, boundaries, ingress, and internal routing described as configuration rather than as accumulated manual state.
Secrets and certificates
Credentials issued, rotated, and revoked through one path, never pasted into the deployment that needs them.
Observability
Metrics, logs, traces, and alerts as a property of the platform, so a service becomes observable by running rather than by being instrumented from scratch.
Delivery is declarative
Desired state lives in version control
What should be running, where, and in which version is a reviewed change to a repository rather than an action someone took at a terminal. The cluster reconciles toward it continuously.
Packaging is separated from placement
How a workload is built and configured is one artifact; which environment receives it is another. That split is what makes promotion, rollback, and multi-environment operation ordinary instead of heroic.
Everything is versioned and reversible
A rollout that cannot be rolled back is an experiment on the community, and it is treated as one.
The estate is data
Physical and virtual assets, sites, racks, machines, addresses, and their relationships are recorded as structured inventory that automation reads and writes. Provisioning, capacity planning, and recovery all draw on that record, which is why the inventory is maintained as a system of record rather than as a diagram someone updates after incidents.
Recovery is a tested property
Backups that have never been restored are a belief. The substrate treats restoration, failover, and rebuild as exercises with results, because the day a community needs them is the day it cannot afford to find out.
Self-hosting is judged by operability
A self-hostable system without update discipline, observability, compatibility tests, and migration tooling offers the appearance of autonomy. The substrate is what makes the offer real, and it is the reason infrastructure is treated as a first-class product rather than as plumbing.