Shipping "kubectl create something" plugin is impossible

[sttts]

Wile looking into shipping a kubectl create something plugin via krew I noticed that the - => _ rewriting in

krew/internal/installation/install.go

Line 242 in 48b3ea1

name = strings.ReplaceAll(name, "-", "_")

makes this impossible right now.

Background: kubectl allows these create plugins (and only those) since kubernetes/kubernetes#124123.

Basically we have this situation:

• $ kubectl create-something maps to the binary kubectl-create_something
• $ kubectl create something maps to the binary kubectl-create-something
• krew does the mentioned rewriting, turning the binary of the create-something plugin into the kubectl-create_something binary on disk (even though the binary in the plugin tar is properly named.

How could we allow to ship a kubectl-create-something binary?

xref https://kubernetes.slack.com/archives/CTJNMDM24/p1728488438493869

[ahmetb]

TIL kubectl create [...] now allows plugins as of v1.30. I think we should have a spec field in the Plugin manifest to facilitate this. Open questions:

1. Do we do this just for kubectl create? what if kubectl changes its mind to allow more builtin commands to support plugins, too?

2. Do we only skip the first dash, or do we preserve all dashes

   For example should a Plugin with name: create-x-y-z be linked as kubectl-create-x_y_z or kubectl-create-x-y-z? The latter technically allows invocations like:

   • kubectl create x-y-z
   • kubectl create x y z
   • kubectl create x-y z
   • ...

   (Krew intentionally only supports the first form, i.e. plugin is registered only with its metadata.name, as-is).

/kind feature
/priority important-soon

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten

[mjudeikis]

/remove-lifecycle rotten

[ahmetb]

/lifecycle frozen

[mjudeikis]

@ahmetb, is there any interest in reopening this? I had proof of concept for this a long time ago, and we discussed it at KubeCon.

If there is, I can try working on this again