0.10.17 — one sum of ratings

The report said: the English version has no star rating. The diagnosis was different from the report. The rating block was in place and taking clicks — it was simply counting votes in that instance’s own database, where there was one. Against a hundred and sixty-two votes cast on the other side, this looked like a missing feature; it was a split pool.

The split had been introduced two hours earlier, deliberately and with a reason: a rating measures the page, so it belongs to a particular translation. The argument does not survive first contact with reality. **A reader does not rate the quality of a translation — they rate the text.** A star under the translation says the same thing as a star under the original.

So the sum is one, counted by the entry key, and the language of the version is recorded with the individual vote: it says who reads which side and does not split the result. A hundred and fifty-two Polish rows and one English row merged into a hundred and fifty-two entries and a hundred and sixty-three votes, with not a single unrecognised one.

The view stays separate, and that is now the whole difference: it alone measures the fact that a particular page was opened by a particular reader. The border drawn in Architecture 27 moved by one position and that entry has been corrected, because it describes the present state. This changelog stays as it is — it describes a release, not a state.