Closed benoit74 closed 1 month ago
After discussing with @rgaudin it is clear that we do not really have a precise idea of how localization should work. So for 4.0.0 we will rather remove i18n translation. It is anyway rarely used by scrapers: ted => barely translated in Hindi, not used in Zimfarm ; openedx => barely translated in Hindi, not working anymore ; nautilus => only translated to FR.
Plan is to confirm what is the proper API and proper way of doing translations across scrapers without current issues on system dependencies.
See https://github.com/openzim/python-scraperlib/issues/188 for the return of translation support in python-scraperlib
Currently, for i18n we rely directly on system locale. We should depend on and use an external locale-data such as what Babel does.
This would provide a real support for Alpine Linux for instance (see https://github.com/openzim/python-scraperlib/pull/133#pullrequestreview-1877241783 for more context / details about Alpine issue) and less system-dependent behaviors.
We have to check all our use cases to check how to implement this with minimal/no impact.