Manish Goregaokar

> We're not regressing. We're not "sacrificing" anything in the more narrow context of this PR Okay, but the "narrow context of the PR" is not the context in which...

> variable length operative term here, I think we'll be applying these to most fixed length ones too and I'm not opposed to the VarULE macros exposing these methods too,...

@sffc I don't think we need to test the invariant at all, as long as the macro controls generating the struct.

Going to move discussion to https://github.com/unicode-org/icu4x/issues/5127

"baked size" here is the size of the .rs file, yes?

Rust already has trait naming conventions: https://rust-lang.github.io/rfcs/0344-conventions-galore.html#trait-naming, which suggests transitive verbs, if not verbs then nouns, if not nouns, only _then_ adjectives. Put differently: - Describe what it does (usually,...

Yeah. ICU4X does not do _that much_ weird stuff that we should need to be a special snowflake in these ways. I'm always a bit wary when I find crates...

> There are cases where we are getting _close_ in `icu_datetime`. For example, `FieldSet` is a struct, but `ConstFieldSet` is a trait; I was slightly more comfortable with the `HasConstFieldSet`...

> On trait names: I generally strongly prefer general solutions that avoid clashes ("just work"), but we can fall back to a case-by-case bikeshed given that the referenced Rust RFC...

I think the Rust trait scheme would have unambiguously given BufferProvide and BlobDataProvide for traits. We didn't follow it, and what we chose instead is still acceptable, but "verbs first,...