Repository navigation
v1.3 planning #1295
Description
We should consider releasing the spec v1.3.0 soon, as the ETA of runc v1.4.0 is approaching
The current items in the v1.3.0 milestone
specs-go/config: add Landlock LSM support #1111Carry #1111: specs-go/config: add Landlock LSM support #1241- Clarify Intel RDT configuration #1196
- config: Add Hardware description object to the VM configuration #1209
- linux: clarify pids cgroup settings #1279
- CI: add codespell, bump golangci-lint #1281
- Add FreeBSD as a platform #1286
- schema: fix json #1297
- ci: use supported Go versions #1300
Probably the landlock ones should be deferred (again) to v1.4.0+, as the JSON schema is not stabilized yet
Other ones can be safely merged soon?
Reacted by Markus Lehtonen, Rodrigo Campos Catelin, Toru Komatsu and h-vetinari
Activity
alice-sowerby commentedon Sep 2, 2025
Hello! I would like to understand the timeline for v1.3 as the FreeBSD work is currently close to completion and we'd like to see if it could possibly be included. If you could share any dates that we can work to, that would be super helpful. Or, if it's too late now, that would also be good to know.
AkihiroSuda commentedon Sep 2, 2025
According to #1286 (comment)
The POC for ocijail seems ready in https://github.com/dfr/ocijail/tree/freebsd-oci-runtime , but the POC for runj doesn't seem ready AFAICS: samuelkarp/runj#60
spec v1.3 should be released before the release of runc v1.4, which is targeted for 2025-10-31:
We have already missed the schedule for runc v1.4-rc1 (2025-08-28), so the release of runc v1.4 may be potentially delayed, though.
dfr commentedon Sep 10, 2025
I know that Sam has was looking into updating runj to work with the proposed FreeBSD extension (#1286) but I don't think he has made any of that work public yet.
As far as I know, at this point, Podman+ocijail is the the most mature option for using OCI style containers on FreeBSD with a healthy group of early adopters using it. The containerd+runj alternative is somewhat behind, which is understandable given that Sam has a lot on his plate right now.
Would it be reasonable to accept a single POC implementation initially with a runj POC available in the near future but potentially after runtime-spec 1.3?
cyphar commentedon Sep 10, 2025
@AkihiroSuda FWIW, I don't plan to delay the runc GA release.
@dfr
I would prefer to not release a PoC included in the spec if we aren't absolutely sure it won't be changed later. We had a lot of work put into the Linux spec over several years and we still made some mistakes that I wish we could easily fix -- multiple implementations may have helped us with that (though some of them were design mistakes, such as how we deal with devices...).
Given that we tend to do runtime-spec releases on an as-needed basis, I'm okay with saying that we can plan to do a release with FreeBSD support once it's ready and merged and so there isn't a rush to make a particular release.
kolyshkin commentedon Oct 10, 2025
Added #1297 to the list (trivial fix).
kolyshkin commentedon Oct 21, 2025
Added #1300 (another trivial one but a net improvement); hope we can merge it while people still review #1286.
alice-sowerby commentedon Oct 28, 2025
Thanks for helping get the FreeBSD PR over the line, much appreciated to all involved!