HIP-0011: Informational Helm CRD HIP

[technosophos]

This is a draft of a document to provide a complete explanation of the CRD conundrum in Helm. The intention is to have one formal statement that we can direct others to so that they understand:

• The nature of the problem
• The reason for Helm's current implementation
• The guidelines for future alternatives
• The time frame for reconsidering our design choices

Signed-off-by: Matt Butcher matt.butcher@microsoft.com

[mattfarina]

In general, I like this write-up. I think every maintainer and contributor to the space should read it.

[thomastaylor312]

💯

[thomastaylor312]

Nit: remove extra space between Global and Resources

[hickeyma]

This is a really good detailed discussion on an area which has proved difficult to explain in the Helm community. It will be helpful when trying to find solutions to people's issues. Thanks @technosophos

[bacongobbler]

Hi @technosophos!

As #140 has now been approved and accepted, would you mind following the process described in HIP 1 and rewrite this as an "informational" HIP?

Let me know if you have any questions about the process. I'd be happy to answer them.

Thank you!

[jtigger]

This is a challenging situation. Not just in where all the various machinery are at, but where we are as a result as a broader community.

That y'all took the time to elaborate all of this accounts to genuine "thought leadership" within this community, in my book.

As a member, I'm grateful. 🙏

[elmariofredo]

Hi,

I have run into issue of helm installing and not updating CRDs and end up here. I have tried to read this explanation deeply and
I have two questions:

1. Can you please provide examples what exactly could go wrong and by examples I mean real world examples not just statement that this can happen as I feel that either I'm living in strange bubble or helm contributors see more into how chart authors are building their charts. If there is small number of rare poorly designed charts is it really worth of these precautions? I think that It's helm chart authors responsibility to build quality chart which will not break users setup not helm same way that developer is responsible for their program and not author of programming language which developer used.

2. If helm really don't want to handle CRD then do not handle them at all not even installing as it doesn't make sense and can lead to erratic behaviour like upgrade old crd and new controller.

Sorry to jump directly to PR but as you have closed related bugs there is no other space to ask my question.

Thank You

[philoserf]

Adding some suggestions on word use, phrasing.

[technosophos]

I was granted HIP approval, but no HIP number was granted. So I am claiming HIP-0011.

[technosophos]

I have reformatted and fixed up this document. Anything I marked as "resolved" I either changed or decided not to further consider the matter. Given that I can't solve the problem, I believe that this HIP is suitable for merging as an informational document. From there, perhaps we can figure out a better way to have the ongoing dialog about this issue--and have a stable reference that others can use.

[bacongobbler]

Typically "informational" HIPs are considered final after approval, as they're meant to serve as information to the general Helm community, and it's not really considered a living document compared to "feature" or "process" HIPs. Do you expect there to be further changes to this document or would you consider this as a final draft?

Do keep in mind that newer HIPs can replace or deprecate older HIPs, so if we change how CRDs are handled later on we can mark this one as superseded.

[technosophos]

The intention was to document why Helm 3 works the way it does. So it is informational IMO. If a new thing happens for Helm 4, I envision a new HIP for that.

The main thing I was hoping to accomplish is to give people a substantial reference to understand what we have already done and why.

[mattfarina]

@bacongobbler @technosophos I think the status can be set to final. Is that the only change being noted here?

[bacongobbler]

yes

[technosophos]

@bacongobbler are we good to merge at this point? You still have a nay vote on this PR.