Fedora is migrating to Weblate, I did a proof of concept demonstrating the changes that Weblate will have regarding our Websites.
See https://pagure.io/fedora-websites/issue/1013 and https://docs.weblate.org/

I created here a manual file to manual pull translation from Weblate.
Here are the impacts and questions of the Weblate migration:

1. No more empty translation files, how much does this impact website generation?
2. As a consequence, LINGUAS won't contain all possible values, how much does this impact website generation and user experience? Typical workflow: I'm Chinese, and all website aren't translated in Chinese. I should be able to always have a Chinese version when it exists, even if I come from a non translated Fedora Website.
3. Why do we have an en.po in our repository? What's the use case?

This pull requests contains po removal and po updates on 100 files, which makes it way to big.

Correct implementation proposal: I suggest to have a dedicated repository for websites translation, and allow Weblate to do commits.

1. Our Websites automation would push new pot files on this repository and pull po files.
2. Weblate will update po files when required and update all LINGUAS files.

If the naming rule is clean, I can do some auto-detection so that a new website to translate is automated by Weblate without requiring any administrative action in the translation platform.

Could you please create a localization repository with @jibecfed as maintainer (I'll have to change project's settings) and @weblatebot with commit rights?
I assume something like https://pagure.io/fedora-web/translations would do the job.