It very well may turn out that the ability to specify devices to be created in a z/OS container may be unnecessary and that the definition should be replaced with more useful properties.

Given that this is speculative, and anything we merge might end up in a release, perhaps this should wait until there's something concrete to include? Alternatively/also, it might be prudent to mark all the bits related to this platform as "experimental" or "in-progress" in some way until it's deemed at least minimally feature complete (if the goal is to merge adjustments incrementally before the full functionality is known/working).

Sorry for the delay. Yes, just as you said, the goal is the merge adjustments incrementally.

Let me explain a bit more about /dev on z/OS. Although the major/minor number of devices are different on z/OS when compared to Linux, they can be created in the same way - mknod. The more interesting thing on z/OS is that the number of different kinds of devices is fairly limited and that they are created by the system either during initialization or on demand. For example, if a request to open /dev/console is done and the file is not there, the system will create a /dev/console with an appropriate major/minor number. It is this on-demand creation that led me to the statement that it is not very useful on z/OS.

Do you think that just an in progress note on the zos line in the list of Platforms in spec.md would satisfy the raised concern?

I added a commit to Indicate work in progress.
It's not the changes I was leaning toward in my previous comment, but I think it covers more:
Added "This document is a work in progress." to config-zos.md.

IBM is looking at providing z/OS systems for CI. 🙂

It's been a while. Is anything else needed for this PR to get approved/merged?

I'd suggest a rebase (to try and fix the CI failure) and a squash (since I don't think we want three commits that together amount to "this is coming"). 😅

(Maybe that will help encourage other members of @opencontainers/runtime-spec-maintainers to opine/review.)

@tianon @samuelkarp Thanks for the suggestions! The 3 commits have been replaced by a single commit that incorporates your suggestions.

Would love to check in on this and see what's the latest status -- this seemed like an effort with a lot of steam behind it three years ago, but we (in the OCI / runtime-spec) haven't really heard anything since. Should we consider removing this temporary/experimental zos from the spec? Should it be updated?

@tianon, thanks for checking in!

Certainly don't want to remove. My team wants to push forward to remove the temporary/experimental status. IBM z/OS Container Platform is available to z/OS customers! We've been waiting for a while for all the pieces to come together, but everything is available now.

There are some changes from this initial PR. I believe it mostly amounts to the addition of namespaces on z/OS. I expect that someone on my team will open a PR in the coming weeks. We will want to remove the devices. If it is felt we need to stage that kind of change, we can discuss in the upcoming PR.