Your opinion: "sr_Latn" or "sr@Latn"

[buhtz]

Hello,
I am interested in your opinion.

Serbian in latin script: Would you choose "sr_Latn" or "sr@Latn"?

Weblate ofers "sr_Latn" but not "sr@Latn".

But the Debian folks (using GNU gettext) telling me I should use "sr@Latn". I am not able to find an official list or standard for such language code suffixes.

Some more background:
Upstream project: https://github.com/bit-team/backintime
Weblate instance: https://translate.codeberg.org/engage/backintime/
Discussion at debian-i18n: https://lists.debian.org/debian-i18n/2025/03/msg00000.html
Debian package: https://tracker.debian.org/pkg/backintime
The gettext error message: https://i18n.debian.org/l10n-pkg-status/b/backintime.html

Regards,
Christian

[nijel]

sr_Latn follows BCP 47 standard, sr@Latn is the legacy one introduced by Gettext before these kinds of tags were standardized. I'm not sure if the GNU gettext library currently supports BCP 47 tags, so you might have to stick with legacy one if using that.

If you name your files sr@Latn, Weblat will map this to sr_Latn for its purposes (but the file name will stay the same). You can read more on that in https://docs.weblate.org/en/latest/admin/languages.html#parsing-language-codes.

[buhtz]

Thank you very much for your feedback.
Files named sr_Latn.mo can be handled by gettext (Debian, Python, Qt, ...) without problems. So technically there is no problem. I just try to understand.

So it could be said, that sr@Latn is kind of "old" and related to a project-specific standard. That standard could be named "outdated" today because it was replaced by BCP 47 which is a real worldwide consensus based standard?

[nijel]

It's not about being able to read such files, but about the whole locales ecosystem in Linux. I have no current knowledge on what is the current-recommended approach there to set Latin or Cyrillic variant of the locale and if that gracefully maps to BCP 47 tags.

PS: I think it was actually sr@latin, not sr@Latn before.