Comment 4 for bug 681872

The Debian maintainers are correct about the ISO639-2 names. But those names are not for selecting languages, they are for classifying languages. A simple way to demonstrate this is to imagine if someone was to translate Debian or Ubuntu into the Blackfoot language, and someone else was to translate it into the Malecite-Passamaquoddy language. Following ISO639-2 to the letter would require them both to be listed as "Algonquian languages", which would be nonsense, because they're mutually unintelligible languages. "Algonquian languages" is a useful classification, but it's a useless identifier.

I would be surprised if there is any software in Ubuntu *or* in Debian that uses iso-codes for classifying languages, rather than for offering language choices. So if iso-codes sticks exactly to ISO639-2, then it is not fit for the purpose of offering language choices, and there needs to be a language-codes package or something to override or replace it.

A much simpler solution, though, would be to recognize that the ISO639-2 list is also internally inconsistent. For example, it has items for "English, Old (ca.450-1100)" and "English, Middle (1100-1500)" -- but it doesn't have "English, Modern (1500-)", it just has "English". Greek should be treated the same way.

The equivalent bug in Launchpad Translations was bug 81158, fixed in 2007.