Regular Release Schedule

[duncanam]

I am currently using cuda-oxide to develop some pretty slick CFD solvers. The main issue I'm running into right now is that it's tricky for me to integrate it in to my monorepo due to releases not quite being regular. I use release-plz to automatically bump changes, and I think I've found that I can most easily integrate it by publishing the cuda-oxide packages to our internal Kellnr under the current release version. However, any updates here require me to go into Kellnr and delete the old package and re-upload a new one due to cargo's packaging immutability rules. Can I request a more regular release cadence, perhaps leveraging the release candidate form for package versions:

<semantic version>-rc.1, e.g. 0.2.2-rc.1

I think this would help adoption as it would make integrating new updates a bit easier.

[duncanam]

To add to this, if we adopt release-plz, we can use this field to automatically tag as prerelease candidates:

https://release-plz.dev/docs/config#the-git_release_type-field

[nihalpasham]

Thanks for raising this, and great to hear cuda-oxide is powering CFD solvers. Would love to hear more about that at some point.

Quick update on where things stand:

- Most updates are landing on main right now: We haven't cut a release since v0.2.1, so main is the branch to track if you want the latest. For your monorepo, a git dependency with a rev on main might save you the Kellnr delete-and-reupload cycle in the meantime.
- We're intentionally holding off on a new release for now: cuda-oxide itself is switching to a monorepo by the end of next month, and cutting a release right before that restructuring would create more churn for downstream users than it solves.
- More details soon: I'll share the monorepo plan as we get closer to that date.
- On rc versions and release-plz: good suggestion, noted. The monorepo switch means we'll be revisiting release tooling anyway, and a prerelease channel like x.y.z-rc.N fits naturally into that. No commitment on the specific tooling yet, but a more predictable cadence is the goal.

[duncanam]

Thanks for the thoughts, looking forward to that. And yeah, really enjoying cuda-oxide. It's currently powering some internal industrial production fluid flow simulators. Just wanted to mention that part because it's definitely used professionally so progress on this project is super appreciated. Hopefully at some point we can share a bit more on how our solvers work.