Manish Goregaokar

> I'll add to this: the fact that it is gated on `compiled_data` means that it is not a universal fixed default! Like, I don't want people's `Default::default()` to break...

> Then two months later they realize that they should disable the compiled_data feature. But doing so causes mem::take to fail to compile. This ... seems like exactly what we...

> The problem with docs on trait impls is that they are not discoverable I mean, if you get a "missing Default impl" when you play with features I don't...

TLDR: I think the proposed policy works but I think we should talk about what we think of as singletons here a bit more ------------- > I think the fundamental...

> it has a new function so clearly there is a default value I think this is a misrepresentation of my position, I don't think you're arguing against something anyone...

In favor of the principle, think the wording needs more work.

I agree, 3 or 2, preference for 3

That's a good argument in favor of 2, I like that. Seems like we're in consensus on 2?

#5650 solves the 2.0 blocker here. The rest is gravy and doable in a backwards compatible way

We've started integrating some of the algorithms via https://github.com/unicode-org/icu4x/pull/6870. As previously discussed, we're focusing on the algorithm improvements in fixed to/from iso code (but not as much the other ones;...