Just saw a supposed roadmap slide from a Helm maintainer's talk. If it's real, Helm 4 is doubling down on complexity we don't need.
The big points from the leak:
* Deprecating `requirements.yaml` entirely, forcing everything into Chart.yaml dependencies.
* New "artifact registry" layer, adding another moving part.
* More Lua hooks for "advanced" use cases.
We're already drowning in YAML and abstractions. This is the wrong direction.
Example of the new forced dependency format they're pushing:
```yaml
dependencies:
- name: nginx
repository: https://new-artifact-registry.corp
version: 1.2.0
condition: nginx.enabled
tags:
- web
```
Now we need a registry *and* tags? This adds zero functional value over a simple, versioned http link.
We should be stripping tools back, not adding more glue. Most of us just need a simple templating engine that pulls in a few k8s manifests. The obsession with building a full-blown "package manager" for k8s is creating more problems than it solves.
If this is true, it's the final push for many to move to plain kustomize or basic gitops pipelines.
Simplicity is the ultimate sophistication
Totally feel you on the complexity creep. The Lua hooks bit especially - most teams I work with are already stretching to debug Helm's own templating, adding another scripting layer sounds like a support nightmare.
But I'm weirdly okay with killing `requirements.yaml` if it means one less file to track. That merge was always clunky. The tags thing, though... unless those tags become actionable selectors in the CLI, it's just metadata for metadata's sake.
Honestly, if this leak is real, my biggest worry is the new artifact registry. We just finished migrating everything to OCI. Another registry change would break our CI pipelines for months. The ROI on that migration would be brutal.
Keep automating!