Krishnan Anantheswaran

@jamestalmage says it best here: https://github.com/istanbuljs/append-transform/issues/8#issuecomment-228942804

> I think this would be true for functions like function-kcl and function-cue that just use the protobufs as their data model too. Things like oxr.resource['name'] would need to be...

I'm going to play devil's advocate with this change for reasons different from @negz 's objections. * This change only works because the child is a k8s resource that can...

Don't disagree with the utility of this change, but based on discussions above this feels like "extra info" that could be attached to a resource rather than an actual child...

If the end-goal is the CLI and only the CLI, any direction is fine. But if the end-goal is that this will be usable as a library at some point...

the only problem is one of maintaining this function that was written for helm2 and make sure it still works for helm3. I'm ok leaving it in if that is...

Why couldn't you do all this in a custom transport? We already do user-agent injection and other stuff like this using a custom transport.

> This might be worth documenting in a blog post, what do you think? That would be awesome!

### A totally bad benchmark This only serves to show that increasing concurrency _does_ matter. Some numbers on my laptop after hard-coding `qps` and `burst` to `50` and `100` such...

Interesting - lint is failing for go 1.22 int range usage but `make reviewable` passed on my laptop...