When a page is served to user by a machine translation service, most probably outside a wikipedia domain. There should be standard way to detect that context.
- It should be translation service independent if possible
- Should be able to identify the source language, target language
- Should be able to identify the machine translation service
This may require some customization at the external services as well.