Consider unifying this repo with upstream

[mattip]

Since the migration to cibuildwheel, it would seem this repo could be merged with the upstream statsmodels. That would get more eyes on the wheel building, and would make it easier to manage things like the API secret by the project maintainers. Is there a reason to keep it separate here?

[bashtage]

We should probably just migrate it to the statsmodels org as a speerate repo. I agree there isn't much of a case to continue to use this org.

[agriyakhetarpal]

I fully agree with this, it took me a short while yesterday to find out where statsmodels was publishing its nightly wheels from! This repository is hidden in the nooks and crannies of a separate organisation :)

[josef-pkt]

numpy, scipy and pandas have already moved.

So, I don't see a reason to move statsmodels builds also.

@bashtage
How should we do this?
for example
https://docs.github.com/en/repositories/creating-and-managing-repositories/transferring-a-repository#transferring-a-repository-owned-by-your-organization
or just forking, or creating a separate repo

AFAIU, you have permission on both organizations to make a transfer.

[agriyakhetarpal]

It could also be moved and maintained as a standalone workflow in the main statsmodels repository, which would avoid the need for updating the submodule before building.

[mattip]

NumPy and SciPy recently moved their release process to a separate repo, mainly for security reasons. They left a small selection of wheels building in the main repo to make sure the source-build from an sdist download still works.

[bashtage]

I agree we should move it. @josef-pkt could you create an empty repo and give me the correct permissions, or let me create the repo.

[mattip]

Ping. Also is the project still using https://anaconda.org/multibuild-wheels-staging/statsmodels ? Most other projects have stopped uploading there, and are using trusted publishing (edit: or downloading the wheels locally) instead.

[bashtage]

Yes. We used it on Friday for 0.14.6. what are others doing with releases?

[mattip]

Scipy is still using the staging area for releases, but is moving to trusted publishing directly from github.

NumPy already uses trusted publishing, as does pandas.

For openblas-libs (scipy-openblas32 and scipy-openblas64) I upload to scientific-python-nightly-wheels since (until very recently) it uses travis and travis does not have trusted publishing.

[bashtage]

We will have to look into that. This said, you could delete all of the statsmodels.wheels there except 0.14.6 since they are all on PYPi.

[mattip]

delete all of the statsmodels.wheels there except 0.14.6

Done