-
Notifications
You must be signed in to change notification settings - Fork 13
How do we propose these changes land across OCI repos? #62
Comments
The OCI Artifacts repo was always intended to evolve to support broader scenarios. We said it was scoped, at the time, but I don't think it was ever said it couldn't evolve, rather it needs TOB approval to become a spec. |
I don't think the intention of the WG was to propose anything to the artifacts repo. Instead, I believe we intend to propose changes to the existing OCI specs (somebody please correct me if I'm wrong). How and whether artifacts evolves toward becoming a spec, or part of a spec, feels separate and orthogonal to me in answering how (or whether) the reference types WG's proposals land in OCI specs. |
The only reason I mention anything besides image-spec/distribution-spec is the new manifest type being proposed in Proposal E: https://github.com/opencontainers/wg-reference-types/blob/main/docs/proposals/PROPOSAL_E.md#artifact-spec Where would that belong? In my opinion, this either
|
or distribution spec, or rename/rescope artifacts, or some other combination... is a good sign the wg is getting closer to a proposal/output :-) |
The simplest option to me seems to be be to keep the new manifest types in image-spec, and perhaps consider renaming it later. It already includes other types such as descriptor and index, so another one isn't that far from where we are today. Then the new API changes would live in distribution-spec. |
Ultimately, this will be a TOB decision. (Perhaps tagging @opencontainers/tob will raise some visibility). I believe we will have changes to image-spec and distribution-spec that will document here and then raise as PR's to this projects. For artifact-spec, we can define what that should look like as a folder within the WG repo. Once we get to that point, the TOB can make the decision on whether that gets pushed as a new repo, artifacts gets renamed and extended with new content, or artifacts gets merged into the new artifacts-spec repo. The other option I thought about for artifact-spec was to create a new repo and have the WG populate it ourselves. That might get more complicated with sorting out maintainers of the new repo just to make the proposal, so I was leaning towards the simple solution of a directory in this repo.
I was pushing for that early on, but it got shot down. I believe the image-spec maintainers want to limit the scope of their repo. |
Do you have a link to that? I don't remember it :( Keeping all the manifest specs together seems the simplest to me. The maintainers have changed, so it might be worth revisiting. |
I'm not sure when exactly that was, but I'm pretty sure it was within the last year. @samuelkarp may remember. |
Thanks for the tag @sudo-bmitch!
I don't really have a memory of this, but possibly opencontainers/image-spec#907 (@vbatts) is relevant here as it implies that the image-spec maintainers think things-that-are-not-images might not belong in the image-spec?
My initial inclination is to agree with this, but (a) I'm not speaking on behalf of the TOB here, (b) I haven't really been following the WGs efforts so I'm not up to date on the details of any proposal, and (c) I'd want to hear both what the WG itself wants to propose as well as what the image-spec maintainers think about it.
💯 I'm excited to see that we're getting closer! |
I for one am on board for a rename of image spec and clarification of the different structures. Since it was brought up recently about moving something like the descriptor from image-spec into distribution-spec, it was pointed out that this would be a breaking change. |
@vbatts we discussed on the call and agree we should go this general direction. Maybe rename image-spec to
?? Either way, we seemed to agree on 2 PRs as the initial output of this working group:
|
Seems like we are on the same page on this, closing |
Came up during the general call today-
Is this a recommendation that should from from the WG or TOB?
cc @SteveLasker @sajayantony @sudo-bmitch @mikebrow
The text was updated successfully, but these errors were encountered: