Manish Goregaokar

I'm fine with this landing as is but in the long run I want to make sure it is incredibly obvious which of these many mappers is right for the...

As long as it's not pulling in more deps I'm fine with it.

I kinda like the Unvalidated prefix and don't feel bad about the negation. Could be DeferredStr instead (`serdefer` makes for a cute crate name in that case). Or `LaterStr` if...

Not opposed.

In this case I can't immediately think of anything and `unvalidated` is available and accurate, we should just use that imo.

I'm fine with `unvalidated::UnvalidatedStr`. I don't really think any of the other names work well here. I think the un is warranted here.

> What did you all think of `PotentiallyIllFormedUtf8` Not strongly opposed, feels long. Still prefer the current `Unvalidated` naming. Not a fan of `pill`. I think it's okayish.

Discussion Conclusions: 1. This lives in a crate named after the type (probably), that depends on `serde` and `zerovec`, with feature flags. 2. Fine with using these for `_utf8` and...

Proposal: 1. Error enums are fine grained, but `#[non_exhaustive]` by default. This will imply some level of error wrapping 2. At a case by case level we can make decisions...

For things with options, one solution we can have is requiring validation if the options ever gain invariants. We're not making the decision for now. Either split options/validatedoptions ```rust= let...