RFR: 8317979: Use TZ database style abbreviations in the CLDR locale provider [v3]

[Steven Loomis]

On Tue, 17 Oct 2023 16:52:12 GMT, Naoto Sato <naoto at openjdk.org> wrote:

>> CLDR provides very few short names for time zones, such as PST/PDT. This will typically end up substituting names from the COMPAT provider. Once the COMPAT is removed, they will be displayed in the GMT format, i.e., GMT+XX:YY. Although some of the short names in the COMPAT provider are somewhat questionable (less common ones are simply made up from the long names by taking the initials), it would not be desirable for them to fall back to the GMT format.
>> To mitigate the situation, CLDR can use the abbreviated names from the TZ database, which contains legacy (major) short names as FORMAT. The CLDR provider can use them instead of the GMT offset style. This enhancement is a precursor to the future removal of the COMPAT provider.
>
> Naoto Sato has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Delay populating GMT format at runtime. Reflecting review comments.

> CLDR provides very few short names for time zones, such as PST/PDT. This will typically end up substituting names from the COMPAT provider. Once the COMPAT is removed, they will be displayed in the GMT format, i.e., GMT+XX:YY. Although some of the short names in the COMPAT provider are somewhat questionable (less common ones are simply made up from the long names by taking the initials), it would not be desirable for them to fall back to the GMT format. To mitigate the situation, CLDR can use the abbreviated names from the TZ database, which contains legacy (major) short names as FORMAT. The CLDR provider can use them instead of the GMT offset style. This enhancement is a precursor to the future removal of the COMPAT provider.

This is intentional, because these short names may not be known to users.  Do you have data that ja-JP, zh-CN, de-DE users expect and are familiar with `PST/PDT`?   This is why CLDR fallback rules would fall back to `Los Angeles Time` for example in the longer name. It's not short, but it's understandable, and the numeric offset can work for short. 

I'd encourage engaging with CLDR-TC to discuss the short names upstream if you have data on the recognizability of these short names.   In fact, CLDR probably has _too many_ short names for zones.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/16206#issuecomment-1767240212