Robert Bastian

The `unvalidated` name is available. We could use `unvalidated::Str` and `unvalidated::Char`

> Can't we come up with some other "clever" term a la elsa::FrozenMap, yoke::Yokeable, ... ? I really *really* dislike all the "clever" names.

I think this is also ok, however it looks harder to reach in an incremental change in 1.5. I also think the defaults table adds a lot of complexity that...

I'd phrase this all more constructively: ```rust pub enum LocaleInclusionMode { /// All requested locales, their ancestors, and their descendants are included, unless /// they fall back to a locale...

Option 2 sounds good. We already have `full` on the CLI, so I think we should continue using that instead of `*`

It depends what you're doing. If we have the langid -> locale family conversion implementations, and you use default options, then the invocation will be ```rust DatagenDriver::new() .with_keys([HelloWorldV1Marker::KEY]) .with_locales_and_fallback(SELECTED_LOCALES, Default::default())...

> Can you clarify how you envision accomplishing this? I thought we had this before, but maybe it's not possible with the current design. I don't think conciseness is the...

Not opposed in general, I'd like to make this decision once we have a final API. I don't want too many methods on the driver, and these don't add any...

https://github.com/unicode-org/icu4x/issues/2639 doesn't really contain any discussion, not even with the people that were linked (I was not). In https://github.com/unicode-org/icu4x/issues/153 we decided this for 1.0, but I'm now convinced this was...

Also, in #153 I did say > @robertbastian - For 1.0 we should at least make sure errors are not overly nested. If there is a data error, there should...