On 2/3/19 8:44 AM, Lyberta wrote:
> Since we introduced char8_t but not char8_t-based streams, early users
> of char8_t will face significant difficulties working with char8_t based
> data. I think we should let iostreams die and think of something simpler
> that uses std::byte and then build text I/O on top of it.
>
> Ideas?

I think adding char8_t (and char16_t and char32_t) support to iostreams
is a separate concern from building something new. I suspect there is
sufficient demand for both. What we're really missing is a proposal for
something new. Well, we're also missing a proposal for how to add
support for char8_t (and char16_t and char32_t) to iostreams. I'm not
sure anyone wants std::u8cout, std::u16cout, and std::u32out. But
std::cout << u8"text" is problematic as well.

In San Diego, SG16 identified support for transcoding interfaces for
program boundaries as one of its highest priorities. So, ideas and
proposals are more than welcome.

Tom.