comradekingu
commented
5 years ago The alternative to per-translator status review is #1203
A binary self-review falls short of a confidence level to issue a suggestion/translation/review by, at ones own ability, where "needs review" is now someone's opinion, or extra even more unknown parties, of a bad source string, badly translated string, possible own or mistakes of others, or research needed. Per-reviewer review serves to avoid mistakes of ones own making, to the extent of ones own knowledge of them.
A second pass is better because the first pass is, in so doing, the third and the fourth is too, albeit for diminishing returns, an improvement. What it does beyond ensuring everyone is incapable of finding flaws, is increase the risk that someone disagrees, and I, for one, really want to know who disagrees, and for what reason.
This presents a venue of opportunity in challenging a lot of people at once. One of them will likely take the time to check up on the change, and as certainty increases, likely revert the change, and maybe even take the time to explain why. Or they will at best all learn at once.
Everyone trusting themselves to translate, or review, should know that unless adding value, not only will someone will see it, it will be fixed. In the best interest of adding value that isn't great, there is no real way to say a string, or project is about 85% right, but 85% right sure is a lot better than not translated. Maybe in the future the string complexion could be rated too, but I digress.
Per review status is where to start, it is the only way to improve quality short of miracles, time nobody has, or perfect auto-translation. Adding to that confidence, even knowing who's reviews are overturned, per account doing it, and/or on the whole, adds some level of confidence.
Messing up the one "needs work", or the one "reviewed"-status, is a disaster, especially so if it isn't malicious, because right now someone needs to see and remember why it was applied right away. I don't think there is an e-mail for checking or unchecking "needs work".
What only first party review does is put strings in the category of "no second party review". That is a good place to start reviewing for anyone else, after doing the "no-review" strings.
Reviewing ones own strings, and having a system to do it, to remember what strings have been reviewed, is also better for the same reason knowing who to trust becomes possible. The premise of making mistakes and trying to avoid them is the same all the way down, so it should be no-one, and at best one person, and (only) by extension, many such persons, and automated systems.
Primarily using roles to forego this is a broken system if quality is the issue. Sometimes projects require review for everything, when the (only) practical alternative is superior.
The single most important thing to ensure quality is knowing what strings are seen, and which ones aren't, more importantly, by whom. I think this would improve quality quite a bit, because it makes it easier to focus without putting a lot of time into keeping track of things, and it allows people that only want to review to be part of the process.