Still, a language monoculture is more convenient for most programmers. As an English-speaking programmer, I'd hate to have to maintain, say, Chinese code.
I think the language constructs/keywords are the "easy" part. But how do you translate identifiers, comments and other "textual" content automatically?
A less hack-y solution might be for new languages to start incorporating more languages in their choices of keywords and class names.
Suppose someone were designing a new language (let's call it Foolang), and these were some of the keywords:
* publico, privado, protegido (access specifiers in Spanish)
* eetha/akhra (if/else in Arabic)
* zhen/jia (true/false in Mandarin)
I think this would be an interesting experiment. It would have to be a compelling language in its own right though.