How are you planning to scale the SE sites to use different languages?

Question

Fast forward a few years, lets say that lots of SE sites will get successful. People will want them in their native language too. They will try to create (for example) Physics.SE in French, German, etc.

What about creating only one site for each topic and something like language tags? It could defaultly load to display English posts and the user would be asked what languages he knows, e.g. English, German, Italian. Only posts in these languages would be shown. The default language could be detected somehow.

Answer 1

For a bit of background, start here: Non-English Sites in Stack Exchange.

While a localization interface has not been developed yet, I believe the various localized versions of proposed sites will end up as separate communities. We could provide a "translate-this-post" interface for one site per subject, but that would overstep one of the critical criteria of creating a successful Stack Exchange community: assuring critical mass.

If you show up to the Astronomy Stack Exchange and select Italian, you would not likely see many answers to your questions. There simply are not enough "experts in Italian" on that site to provide a critical mass of users.

One of the top criteria of creating a Stack Exchange community is that users have a reasonable expectation that they will get an answer to their question. When questions stop getting answers, you create a bad user experience and ultimately hurt the site.

By applying the same critical-mass criteria to ALL communities — whether they are for different subjects or different languages — the critical community-building activities can be preserved. If the Italian community would like a site about astronomy or bicycles or cooking, they'll have to build support and create one. Once created, they can operate the community in a way comfortable to that community, not as some US-imported equivalent of what we think they want to talk about.

These issues are FAR from decided. Your suggestion makes sense. I can only provide a vague account of where we're generally heading at the moment (example). This is a problem for another day.