Inside a Container, Outside the Deposit
Escrow deposit contents for containerised and cloud-native applications
Escrow for cloud-native applications must capture a system assembled from many independently built parts: microservices, container images, orchestration manifests and shared libraries. The traditional deposit of one repository and a build guide assumes a monolith. For a system of forty services, it may capture the code while missing what makes the parts work together.
Why do cloud-native systems challenge traditional deposits?
Cloud-native systems challenge traditional deposits because much of what the system needs lives outside the source code. Builds are spread across many repositories. Base images come from public registries that change or remove versions. Orchestration configuration, private package registries and the delivery pipeline itself are all required to reproduce production, and none of them sits in an application repository by default.
What should a cloud-native deposit contain?
A complete cloud-native deposit reproduces the release, not merely its code:
Component | Deposit requirement |
Service source code | Every repository in the release, with a manifest mapping each service to its commit hash |
Container images | Dockerfiles, plus base images or their digests and a stored copy |
Orchestration | Kubernetes manifests, Helm charts and per-environment values files |
Internal packages | Copies of private libraries, not references to an internal registry |
Pipelines | The CI/CD definitions that build and deploy the system |
Architecture | A service map and the interface specifications between services |
The release manifest is the component most often missing and the most useful to a party rebuilding the system, because it records which version of each service ran together.
How is a cloud-native deposit verified?
Verification of a cloud-native deposit should deploy the whole system to a clean cluster and run smoke tests across service boundaries. Compiling each service separately proves little about whether the system works. Comparing deposited commit hashes and image digests against the production release manifest confirms that the deposit matches what is actually running.
Quick answers
Do public base images need to be deposited? Yes. Registries can remove tags, and a digest alone does not guarantee future availability.
Is a monorepo easier to escrow? It simplifies source completeness, but images, manifests and pipelines are still required.



