For what it’s worth, Christoph insists that I was wrong to raise the alarm. Apparently he did consider Switzerland, as I should’ve been able to tell from the SQL code and demo – a few screenshots of renders of hypothetical data elsewhere in the world. That may have been true to some extent when he issued his proposal. Guessing from his screenshots:
- Switzerland would’ve been labeled “Schweiz ⏎ Svizzera ⏎ Suisse ⏎ Svizra” (separated by newlines) based on
default_language=dename:de=*,name:co=*,name:fr=*, andname:rm=*, respectively. Corsican matters here because its code comes early in the English alphabet. - Valais/Wallis would’ve been labeled “Valais ⏎ Wallis” based on
default_language=frname:fr=*andname:de=*, or maybe “Valais/Wallis” (verbatim) based onname:en=*, not sure. - Grisons would’ve been labeled as “Graubünden ⏎ Grischun ⏎ Grigioni” based on an inherited
default_language=dename:de=*,name:br=*, andname:it=*. Breton’s code comes early in the English alphabet. - Brig and Biel/Bienne would’ve gone unlabeled for lack of the required tags.
- Unsure about Lake Morat and the ferry lines across it.
Since then, some notable changes have taken place in Switzerland. Switzerland itself has default_language=la, which would cause it to be labeled “Confoederatio Helvetica” and Grisons and much inside to go unlabeled. There is talk of removing default_language=*, which would cause the country to go unlabeled too. Or maybe I missed something deep in the thicket of the RfC.
Regardless, as the discussion on GitHub has globalized, it has highlighted cases that would need more thought in other parts of the world that aren’t so tightly bound to political geography. Much of it rehashes examples we’ve already discussed in this thread. I believe it’s just the tip of the iceberg.
In fairness, a map designer should be allowed to label features however they want. We wouldn’t be such a successful project by enforcing our own arbitrary presentational standards on data consumers. (Maybe someone can help me find the novelty style that consists of nothing but labels of feature types.)
On the other hand, the proposed design selectively hides labels that currently render, based on an unintuitive set of heuristics, as a sort of validation rule. That strikes me as a regression and a power move, especially given the proponent’s rhetoric pitting OSM’s supposed Eastern and Western stereotypes against each other. If the community abides it, I hope it will extend the same grace to data consumers that take a gentler, more straightforward approach.