Locale of constructed object design

[sffc]

Split from #173

The design doc submitted in #1237 does not appear to address one of the original questions from #173, which is what types of getters we should expose on formatters to describe what locale they represent.

CC @mihnita

[mihnita]

I think that #173 mixes 2 concerns:

1. The getters for Requested / Valid / Actual locale
2. The first comment in Locale of constructed object & language negotiation in data loading #173 (from Zibi) about Requested / Available / Last Fallback / Resolved Locales

The second one is the one addressed by the "Design doc: locale fallback and negotiation #1237"

The first one I remember being discussed in detail in a meeting.
The various arguments are captured in an issue, or in some meeting notes.
I will search it and will link to it, if I find it.

But the gist of my argument there was that I think that "requested" is OK and possible to implement, but "valid" and " " are basically impossible (at least impossible to give something useful).

To format something we often assemble pieces from a lot of places.
Think "2-23 dollars"
Mixes in numbering formatting symbols (separators, grouping rules), digit types, translation for "dollars", plural rules, currency range format pattern.
And (if for example the requested locale was en-US) these might come from root, en, en-US, Latn (en -> likely subtags).
For dates the first day of the week / week-end come from the country, not the full locale.

Something like actual or valid would return (maybe) a map with info for each piece.
Which is (I would argue) both ugly and useless.

But this does not capture the comments other had.
I will try to find the meeting notes.

[sffc]

Note that we need to answer this question for the purposes of ECMA-402 resolvedOptions.

[sffc]

Discuss in the ICU-TC call.

[sffc]

This is mostly the same as #3906

[sffc]

Closing as duplicate of #3906