Shane F. Carr

https://github.com/servo/rust-harfbuzz/pull/197 is merged. We should be able to remove most or all of the unsafe FFI bindings on the icu4x side once the crate is released.

Waiting for upstream to make a release: https://github.com/servo/rust-harfbuzz/pull/245

We really should support `-u-rg`; it's not a low priority feature. It directly impacts i18n correctness. One way to support both this and `-u-sd` would be for `LocaleFallbacker` to take...

So for `-u-rg` I think we can still emit `LanguageIdentifer` but it's not clear to me how we'd do that for `-u-sd` unless we changed the data model. Ideally we...

I don't think it's worth over-engineering this to avoid a validation which is essentially free. CC @echeran

@hsivonen Thoughts on the latest comments from myself and @echeran?

Ping @hsivonen

Ok, seems reasonable. In any case I'd prefer to fix this in the harfbuzz glue code in the upstream repo so that we avoid interacting with raw harfbuzz types in...

We're competing against a narrow home-grown solution that was written before ICU4X in a time when ICU4C was too big to fit in the resource constrained environment.

ZeroTrie issue: https://github.com/unicode-org/icu4x/issues/4249