search

WeblateOrg / weblate

Web based localization tool with tight version control integration.
https://weblate.org/
GNU General Public License v3.0
4.56k stars 1.01k forks source link

Explicitely warn using non-practical language codes #3559

Closed PanderMusubi closed 3 years ago

PanderMusubi commented 4 years ago

Is your feature request related to a problem? Please describe.

We have discussed this shortly at FOSDEM 2020. The same Dutch language is used in the Netherlands, Belgium and Suriname. Of course, there are local variations, dialects and customs, but in general for software, only one Dutch translation is made. The official rules on spelling, grammar, etc. are identical.

However, the following language codes and locales exist:

  1. nl Dutch in general
  2. nl_AN Dutch in Netherlands Antilles (was a constituent country with the Kingdom of the Netherlands)
  3. nl_AW Dutch in Aruba (is a constituent country with the Kingdom of the Netherlands)
  4. nl_BE Dutch in the Kingdom of Belgium
  5. nl_NL Dutch in the Kingdom of the Netherlands
  6. nl_SR Dutch in the Republic of Suriname

Many many times, developers and translators are not aware that only one Dutch translation needs to be made and/or think that they should choose the language code as specific as possible. Usually this is done with the best of intentions.

In the past en present, this resulted in often two or three translations all trying to do the same translation. Consolidating is sometimes difficult as personal style or choice of words can differ of the different translations, and not that the official language differs. In short, it creates unintentionally a lot of extra work in translating and managing a translation and subsequent software release where the translation is used.

Describe the solution you'd like

Proper internationalized and localized software will switch to e.g. a language code nl if a more specific version such as nl_NL or nl_BE is not available but requested by the operating system.

New and existing projects on Weblate should warn translation maintainers when a language code of nl_* is used and motivate them to refactor and consolidate translations to nl only.

A similar case could be made for en_NZ and en_IE representing English in New Zealand and Ireland with respect to en_GB English in Great Britain. But of the 20 country codes related to English, this is more tricky to define with American, etc.

Of course, the translation maintainer should be able to ignore this warning, but should also be able to reread it at a later point in time somewhere in the settings of a project.

Describe alternatives you've considered

Simply not allowing translations to nl_*, only nl, but that is too forceful and unwanted because some exceptions on for example texts on national law can be found, but these are an extremely tiny amount of the translations. Usually exceptions are found in software content and not in the translation of the software buttons, labels, etc.

Additional context

Implementation could be via these configurations:

Also language code formats such as ca_ES@valencia and ca_ES-valencia should be supported.

nijel commented 4 years ago

I think this should be used to limit list of languages offered to users while adding new translation to reduce amount of garbage added by contributors (this includes adding of en_devel and similar testing locales). On the other side, I don't think Weblate should warn in case such files are already present as there might be good reasons for that.

nijel commented 3 years ago

Just received another feedback on this topic:

What would be amazing, would be custom language lists only shows to users when adding new languages. That way we could have the default language list with all available Weblate languages. And a custom list configured in the Django admin interface only visible for users when adding new languages to a component. In this custom list we could add only the wanted languages for Kodi.

github-actions[bot] commented 3 years ago

Thank you for your report, the issue you have reported has just been fixed.