On Tue, 9 Dec 2025 at 08:26, Nikl Kelbon <kelbonage@gmail.com> wrote:
>  This is not motivation for changing it, you're still just saying it's possible. 

Did you read the code? Or how it may be motivated other than code examples?


I would like to see discussion of why the invalidation rule is there in the first place, with discussion of whether that was a good reason, and whether it still applies now.

The current wording was adding for C++20 by https://cplusplus.github.io/LWG/issue3077 but doesn't actually give rationale for the behaviour, it just re-confirms it and clarifies the wording.