Shane F. Carr

I haven't expressed my opinion yet. I have a preference for try_ naming because that seems to be the newer convention.

Choices: 1. `from_bytes -> Result` 2. `try_from_bytes -> Result` 3. `try_from_bytes -> Result` and `from_bytes (panics)` - @Manishearth - I don't think we should have functions that choose to panic...

Now that I'm in the middle of implementing this, I actually really like the `try_from` convention because it makes method chaining a lot easier to read. We often do `Foo::try_from_bytes().map(...).map_err(...)?`...

@hsivonen What would be an appropriate milestone for this issue?

I am hopeful that Greg's "modern components" work will allow us to handle 11/1223/24 swapping better. Otherwise, we would need to do the expensive "pattern->skeleton->swap->pattern" dance. What I had in...

Some additional thoughts on this: CPT performance is especially critical for Segmenter. Prior to #1839, Segmenter was using a fully expanded table covering almost all code points (1 byte per...

Probably how I'd like to do this would be to introduce something like `CodePointTrieWithStore where S: Deref` and then define `type CodePointTrie = CodePointTrieWithStore`

Discussion: - @sffc - Is this something we can do post 1.0? - @Manishearth - If we typedef it? Yes... it should work, but it's a corner of Rust that...

@hsivonen What would be an appropriate milestone for this issue?

I asked @ptomato for the paper trail. My question is more about whether we should use Plain as a value prop of being consistent with Temporal.