People complain about lost translations again, for example https://www.mediawiki.org/wiki/Topic:Soykmh0gsdacf99j
It would be useful if the software could guard itself against this by checking whether a translation that is being saved internally is significantly shorter than a previously saved version of the same translation. A few deleted words are probably not important, but a whole deleted paragraph is worrying, and several deleted paragraphs or a completely disappeared text is very likely data loss.
This could be used for:
- prevention of data loss (just don't save this new problematic version and don't overwrite a previous good one)
- at least, measuring how often does it happen (sometimes people report it and sometimes they don't), and setting a goal of reducing or eliminating the phenomena
It's challenging, but possible and important.