Open dbrgn opened 4 years ago
I agree that we should state this in a more generic way. I don't believe it's us that must specify how someone implementing a client should detect the system language. Just saying it must do it should be enough.
Ah, thanks for opening this issue! Indeed, this is a problem. I think we should still include at least some guidance on the environment variable for POSIX systems though, because I would guess that implementers writing POSIX-compatible clients will find it helpful.
The spec says:
Windows systems do have environment variables, but the language detection does not work that way. (I don't know right now how detection should be done, but it's not through env vars.)
I would suggest to make the wording a bit more generic, stating that a client must implement reasonable detection of the system language. Then, I would move the parts about LANG, LANGUAGE and LC_MESSAGES to a section that only applies to POSIX-ish systems.