On Wed, 10 Jul 2019 at 22:26, Tom Honermann <tom_at_[hidden]> wrote:

> The situation arises if the implementation assumes imported headers are importable (aka import <foo> implies foo is importable).
> Are we saying this is not a good implementation strategy?
>
> I would say that such an implementation strategy is not valid given the existing wording (and my understanding of the intent of the wording from various discussions).

I disagree with that. The set of importable headers is implementation
defined to be "every header everywhere" in that strategy, and the
wording doesn't
disallow it.