David Arnold

I'm comparing kubenix + nix to `kpt` + `kustomize`. quick notes to self (eventually expanding): - `kpt` → `inputs.X.url = ...?/dir=path/to/deploydir` - `kustomize` → `nix` It would be appropriate for...

@offlinehacker Would you be willing to cede ownership/maintenance so somebody else can take this over and develop it into the future. The future is bright (nickel!)

@takeda my conclusion is: don't use the module system as config validator. Use jsonschema validation. Much more interoperable. Unfortunately, the upstream nix patch to add jsonschema support hasn't been accepted.

This sounds perfect! We should however maintain the `--help` contract for direct invocation.

The relevant contract here is that `help` is not null on any derivation: https://github.com/fluidattacks/makes/blob/d4a5025ab4ae17a498f118bf26dece6b77867e35/src/args/make-derivation/default.nix#L20

> plus a flag for disabling adding such help to the CLI `m . /path/to/script -- --help` This is how the esacpe hatch is currently implemented. It was the best...

I'm prototyping a nomad job tree, so that: ``` /deploy/namespace1/job1 /deploy/namespace1/job2 ... /deploy/namespace2/... ... ``` - jobs ~= `k8s.io/Deployment` - namespace: `dev` / `staging` / `prod`

Btw, I like `repo@branch` better than `repo/branch`. Ideally we could place an RFC upstream to also accept this notation.

I'd like to place 5 :rocket: — one per bullet point :smile:

- I think the best way is to make a "near-perfect" compat layer. - A micro-change switch to `nix3` would imply a trampoline to install a pinned version of `nix3`...