Weird Uuid::default() implementation

[tyilo]

The current Default implementation for Uuid returns Uuid::new() which returns a random UUID v4.

This is different from uuid::Uuid::default() which returns the nil UUID, which seems like a more reasonable default. I would normally expect that T::default() == T::default().

I think Uuid::default() should be changed to match uuid::Uuid and return a nil UUID.

[abr-egn]

Agreed that it's weird, and that the nil UUID would be a better choice for return value. Unfortunately, simply changing it seems like it'd be at high risk of silently breaking existing code in unpleasant ways (e.g. data loss) given the way default can be implicitly used in a lot of contexts to provide fresh values.

I've created RUST-2218 to track potentially including this in the next major version.

[abr-egn]

We discussed this in the team and we're not sure there's enough value here compared to the risk to make this change, even at a major version boundary; if we were to do something like this we'd want to remove the Default impl entirely in 4.0 to make it a compile-breaking change and then re-introduce it in 4.1 with the new behavior.

Do you have a specific use case in mind for this?

[tyilo]

Do you have a specific use case in mind for this?

No, I was just reviewing some code using Uuid::default() and got confused why it worked, as I knew that uuid::Uuid::default() returns the nil UUID.