Em sex., 13 de mai. de 2022 às 10:19, Gabriel Dos Reis
<gdr_at_[hidden]> escreveu:
> The only time there might be any portability concern around "importable"
> is when you have header units. I am suggesting to continue to defer
> to implementation in a way that is even more practical.

Ok, I see what you mean.

I would argue, however, that the portability issues with header units
will be bigger than what we expect if we don't define this more
carefully -- which I think the modules ecosystem tr can do.

Specifically, I would support the notion that library authors have to
explicitly choose to make a header "importable". As opposed to the
notion that "anything you could include, you could now import".

daniel