Avoid low-quality translations

[dontcallmedom]

The previous translation workflow did not have specific quality review in place, and instead relied on marking links to translations with nofollow to discourage abuse for SEO benefits. In the best of cases, one could hope other translators watching the w3c-translators mailing lists would catch bad translations, but there was no formal process or incitation to that effect.

This means

• there is no basic guarantees about the quality of translation (minimal readability, or even rough matching between the translated content and the original document in cases of languages when nobody on the staff could give a quick review)
• there is no opportunity to enforce technical quality or administrative checks on translated documents (e.g. HTML validity, presence of a disclaimer, usage of a given template)
• as a result, there is no clear path for giving more visibility and recognition to the vast majority of translations of good quality

[dontcallmedom]

The proposed new approach would consist of:

• bringing translation as pull request to a target repo for the given spec / language
• upon submission, a small set of per-language reviewers would be notified and be expected to bring a high level review of the translation quality
• in parallel, the submission would trigger a set of automated checks (e.g. using Travis CI)
• once both processes succeed (i.e. at least one thumb up from reviewer, no failure from automated check), the pull request gets (automatically?) merged and the translations get published / linked from W3C site