iso639.json is too limited

[peaceful-james]

The ISO649 language codes do not account for region variants.

For example, "pt-BR" and "pt-PT" (Portuguese in Brazil and Portugal) and both valid locales.

Both are even recognized by the DeepL API (https://developers.deepl.com/docs/resources/supported-languages) (as PT-BR and PT-PT).

Moreover, I believe many of the hardcoded strings in priv/iso639.json could be calculated using functions from different ex_cldr libraries. I am not 100% certain of this but I do know that with the current model where locales are reduced to ISO649 codes, I can not use Kanta to manage Brazil and Portugal variants of Portuguese.

I will branch off my existing PR to hack on this for a while.

[peaceful-james]

Using Cldr, we can get the territory name and the territory "local name". We can also get the unicode flag.

I think it is worth sacrificing these fields from locale to be to correctly support, e.g. pt_BR:

    field :iso639_code, :string
    field :family, :string
    field :wiki_url, :string
    field :colors, {:array, :string}

and introducing these fields:

   field :locale_code, :string
   field :unicode_flag, :string

I can try to work on this but I am concerned about the gettext fork being cancelled. Are there plans to deal with that?

[peaceful-james]

You can assign me to this issue if you want. I might have time over Christmas to give it a go.