On Thu, May 6, 2021 at 8:56 PM Thiago Macieira <thiago_at_[hidden]> wrote:

> On Thursday, 6 May 2021 08:57:44 PDT Corentin Jabot wrote:
> > So the solution is to do exactly what Qt does, put the flag in the build
> > system!
>
> Sure, but what flag is that? The standard can't mandate what flag it is.
> It
> doesn't even acknowledge the existence of flags.
>
> As it is, IBM can comply with the paper and later the standard but still
> not
> make UTF-8 available for our use on EBCDIC machines. "The compiler does
> support it when we tested it, we just didn't give you a flag to use the
> feautre"
>

"implementation-defined" not only mandates that the feature exists and is
available, but also it has to be documented.
If you the user can't compile a utf-8 file with this paper, you were given
a non-conforming compiler.


>
> > We also hope that this will send a signal to users "you should really
> > consider using utf-8" !
>
> Sure, but the point is that with the last mile missing, the feature is not
> useful. If I can't *use* UTF-8, it really makes no difference that the
> compiler supports it. You could mandate that every developer have a rubber
> duck at home in order to write standards-compliant C++, but unless you
> come to
> audit my home, how would you know that my code is compliant?
>
> What I want is to share my file with anyone and know that they will
> compile on
> their system, with their compiler and get the string literals that I had
> meant
> for them to get. I guess this is the "Tom is working on some kind of
> mechanism
> to put in the source file what the encoding of that file is" part of your
> email.
>
> /Divide et conquera/ and all, but I'd like to see that other part.
>
> --
> Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
> Software Architect - Intel DPG Cloud Engineering
>
>
>
>