Consider using script fallback for time zone names and maybe others

[sffc]

The data size optimization landed in #5751 reduces data size in English and other Latin-script languages, but it has minimal impact on other languages.

To reduce this bias, should we consider having a und-Script locale for time zone names. It can contain the names in the language according to likely subtags. This may help languages that share a script, such as sr/uk/ru or zh/zh-Hant/yue. Spot-checking, there are definitely display name overlaps in those languages.

@robertbastian

[robertbastian]

There might be political issues around region display names that need the region to be resolved (and time zone names are often region names).

[robertbastian]

It looks like no script-sharing language overlaps on all names.

[robertbastian]

Ah this makes sense of course, the payloads would have already been deduplicated before. I'm not sure I understand you proposition. We do fallback from non-location names to location names in the same language. What fallback exactly are you suggesting?

[sffc]

Right now we load two payloads, one for the language and one for und. I was suggesting that we instead load two payloads, one for the language and one for the script: und-Latn for en, fr, vi, ..., und-Hani for zh, yue, zh-Hant, ..., und-Cyrl for sr, ru, uk, ...

[sffc]

PR #5960 is not quite what I had in mind. What I had in mind was that, for example:

1. und-Cyrl contains everything from ru, chosen because ru is the likely subtag language for und-Cyrl
2. ru is empty and falls back to und-Cyrl
3. uk gets smaller and contains only those names that differ from und-Cyrl == ru
4. sr gets smaller and contains only those names that differ from und-Cyrl == ru

Just like how we made und contain English data and then en variants are smaller.

[robertbastian]

Ah you mean the deduplication from #5759, not #5751