Hi All, Following up on this discussion: Albert, Stefano and I are organizing a BoF session on vhost-user for Linux Plumbers Conference, so this thread is very timely.
> I get a sense that this is about politics in the end. Do people feel > they are not represented and would like to have more influence in the > vhost-user spec? I think the core issue is that vhost-user has grown beyond just QEMU. The spec is now implemented by rust-vmm, cloud-hypervisor, crosvm, libkrun, and others. All of these projects depend on the spec as their source of truth for interoperability. There are also concrete technical friction points. Let me share some examples from rust-vmm: rust-vmm SHMEM PR https://github.com/rust-vmm/vhost/pull/251 We had to wait for QEMU spec acceptance before merging the implementation. The spec update was delayed with the comment "should get picked for the next round" after QEMU freeze. rust-vmm PR https://github.com/rust-vmm/vhost/pull/339 : was kept as draft "waiting for QEMU spec changes to be finalized." It was eventually merged when they decided the spec was "stable enough" even though QEMU hadn't finalized it yet, which felt awkward. When multiple independent implementations wait on one project's release cycles, it raises the question of whether the governance model matches the ecosystem reality. > Why is the current system not neutral and how will the new system solve > that? For example the virtio-spec has OASIS governance where changes are proposed and merged independently of any implementation's release cycle. Whether that's the right model for vhost-user is worth discussing. > You bring up project neutrality and a model where Michael is no longer > the sole maintainer. It will be necessary to propose a concrete roadmap > and also to explain the concerns about neutrality more so it's clear > they won't be an issue anymore in the future system. You're right that we need a concrete roadmap before making changes. More reason to discuss in the BoF - to work this out with the people actually doing the work: you, Michael, Albert, and others who've been deeply involved. Maybe in the end the solution might be to "improve QEMU's process" (like Michael's submodule suggestion) rather than "move the spec." or something else? that's what we need to figure out together. Would folks here be interested in joining the discussion at Plumbers? BR, Dorinda. On Mon, Jun 1, 2026 at 10:05 PM Michael S. Tsirkin <[email protected]> wrote: > On Mon, Jun 01, 2026 at 11:18:15AM -0400, Stefan Hajnoczi wrote: > > On Mon, Jun 01, 2026 at 04:27:46PM +0200, Albert Esteve wrote: > > > On Mon, Jun 1, 2026 at 3:51 PM Stefan Hajnoczi <[email protected]> > wrote: > > > > > > > > On Mon, Jun 1, 2026 at 9:12 AM Michael S. Tsirkin <[email protected]> > wrote: > > > > > > > > > > On Mon, Jun 01, 2026 at 03:05:50PM +0200, Albert Esteve wrote: > > > > > > On Mon, Jun 1, 2026 at 2:39 PM Michael S. Tsirkin < > [email protected]> wrote: > > > > > > > > > > > > > > On Mon, Jun 01, 2026 at 02:32:11PM +0200, Albert Esteve wrote: > > > > > > > > But also because, in my opinion, separating > > > > > > > > the specification would improve development agility by > decoupling > > > > > > > > specification development from QEMU's review and release > cycles. > > > > > > > > > > > > > > Generally for QEMU this will be less agility, unless I > misunderstand > > > > > > > what is proposed) > > > > > > > > > > > > > > Because presumably there will need to be spec releases then? > > > > > > > > > > > > > > So > > > > > > > new feature -> spec tree -> spec release -> qemu > implementation -> qemu release > > > > > > > > > > > > > > is surely longer that what we have now. > > > > > > > > > > > > I see your point. > > > > > > > > > > > > However, we do not really need to introduce a heavy release > management > > > > > > layer. We could just operate it as a living document, where the > main > > > > > > branch is the authoritative source of truth. > > > > > > > > > > > > For the workflow, development doesn't have to be strictly > sequential > > > > > > either. A contributor can propose the spec update while working > on the > > > > > > implementation, much like we do for VirtIO updates. Actually, > this way > > > > > > one update/change supports the other. > > > > > > > > > > > > I guess my point is that a dedicated repository could lower the > > > > > > barrier for new changes AND keep QEMU's own development speed > mostly > > > > > > unaffected. > > > > > > > > > > > > BR, > > > > > > Albert > > > > > > > > > > Something something submodule? Possibly. If you want to make > progress > > > > > on this, pls think of the process, try it out. > > > > > > > > If I understand correctly, the motivation for moving the spec > > > > somewhere else is to replace the email patch review process with a > git > > > > forge review process? > > > > > > > > This seems like a superficial change and is not worth in my opinion > if > > > > you consider we'll have to redirect from the old spec location and > > > > move the community over to the new repo. > > > > > > The core issue from my perspective is project neutrality and > > > decoupling lifecycles. > > > > > > Currently, protocol updates are tied to QEMU's tree rules and release > > > freezes for example. Since these changes also affect other projects > > > (like rust-vmm, crosvm, DPDK, etc.), separating them may make sense > > > and could streamline the process. > > > > I don't see a reason to block vhost-user.rst changes during QEMU's > > freeze. If Michael sent a vhost-user spec pull request during freeze I > > would merge it. > > > > > >We could actually lose spec > > > > change reviewers in the process either because they don't know what's > > > > going on or decide that they prefer to spend time elsewhere when > faced > > > > with switching processes (the people who review and participate in > > > > discussion do so out of personal interest and as far as I'm aware no > > > > one is employed to work on vhost-user as their #1 priority). > > > > > > This is a fair concern. I hope we maintain reviewer engagement. But it > > > could also be argued that contributors from other communities may be > > > more comfortable with a dedicated project-neutral home. It could well > > > go both ways, but it also represents an opportunity to grow. > > > > > > > > > > > Having said that, here is what I imagine it would involve: > > > > > > > > 1. Michael creates a new repo on a git forge (if he wants it to be > > > > under the GitLab qemu-project organization I can help with creating > > > > the repo, otherwise he creates a new organization and repo). > > > > 2. Discussion happens in Issues. > > > > 3. Spec changes are proposed in Merge Requests. Michael merges them > > > > once consensus has been reached. > > > > > > Mostly yes, though we wouldn't necessarily need Michael as the sole > > > gatekeeper. We could invite co-maintainers from other key projects to > > > share the reviewer load. > > > > > > Also I'd want to clarify that although I am advocating for this > > > change, I do not claim to have the definitive roadmap. I am just > > > sharing my view on why this could be a positive evolution. Consensus > > > may end up being to remain with the current status quo, or any of the > > > other options proposed by Alex. > > > > I get a sense that this is about politics in the end. Do people feel > > they are not represented and would like to have more influence in the > > vhost-user spec? > > > > You bring up project neutrality and a model where Michael is no longer > > the sole maintainer. It will be necessary to propose a concrete roadmap > > and also to explain the concerns about neutrality more so it's clear > > they won't be an issue anymore in the future system. > > > > Why is the current system not neutral and how will the new system solve > > that? > > > > Who should be a co-maintainer? > > > > Stefan > > And also, qemu currently acts like a reference implementation. > Which is something. > > -- > MST > >
