| Shizhao | |
| Nov 7 2018, 2:58 AM |
| F27087485: Screenshot_2018-11-07 (1) 2018年万州公交车坠江事故 - 维基百科，自由的百科全书.png | |
| Nov 7 2018, 3:00 AM |
URLs containing “zh-*” have the same problem
Label lang the correct links:
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Jgiannelos | T230013 LanguagePicker's handling of script suffixes is broken | |||
| Resolved | None | T208927 Label lang error in map |
Seems to be the same as T195318. 'zh-cn' falls back to 'zh-hans' which in turn doesn't fall back to anything. Since South Korea label node (249399362) doesn't have 'name:<lang>' key for neither code then 'name' key value (Korean name) or maybe no name label at all should be used, but instead name in some unexpected language is used, in this case Romansh name ('name:rm' key value).
If find this code which is supposed to do some latinization, but it doesn't seem to work as described. Language using non-Latin script seems to be latinized here. Whilist for Latin-script language 'zh_pinyin' name variants are not shown (e.g. Finnish here), and for Szeged (task referred above) local name ('name' key) is already Latin-script and it shouldn't be necessary to pick up specific 'Latn' suffix.
In addition, as long as internationalization relies on OSM tags then you might want to rework fallbacks so that they wouldn't fall back to codes that some OSM contributors are dedicated to avoid and remove as 'Wikipedia inventions'. So maybe 'zh-hans' could fall back to 'zh', not the other way around.
Update: it seems that now after patch by @Mooeypoo has been deployed it no longer displays "South Korea" in Romansh (rm) for "zh-cn". Now it displays Korean name as "zh-cn" doesn't fall back to "zh". "Japan" however is still latinized as "Nippon" (ja_rm) for "zh-cn". Also, "ko-Latn" value is used to latinize "zh-cn", e.g. see "Namp'o" here.
May I suggest a simple fix: if latinization is too complicated and it's hard to get it work right, so that it wouldn't do more harm than good, then for a start, let's omit it in language picker.
In zoom level 6: Korean displays as “Corea dal sid”, Japan displays as "Nippon"(?), China displays as "China"
In zoom level 4, 5: Korean displays as a Korean language, Japan displays as "Giapun", China displays as "China"
In zoom level 3: Korean displays as a Korean language, Japan displays as "Giapun", China displays as Chinese Hanzi "中国"
I don't know, maybe some tiles for some zoom levels are cached somewhere. "China", "Corea dal sid" and "Giapun" are Romansh name variants that should no longer appear on new tiles. Other issues seem to remain, though.
Fallbacks are not good enough as a solution, The reason is that it is an attempt to fix something by a global algorithm that will have too many local exceptions. E.g. if my language - no - should add a fallback to English we would introduce some very visible errors like Florenca, Venice, Vienna, as well as some far less obvious ones in Russia etc.
These exceptions cannot be fixed easily by us, as the OSM projects view on 'Wikipedia inventions'. However, we now have 1.5 million Wikidata-tags in the OSM database. We should utilize these together with the labels within Wikidata. Add these on top of the language names from OSM, and although there surely will still be errors, at least the most visible of these errors are easily fixed by us.
I'm closing this as language fallbacks have been adjusted as a workaround, and example given in task description apparently displays in correct language now. Underlying LanguagePicker issues are covered in parent task.