On 12/07/2021 09.57, Corentin Jabot wrote:
>
>
> On Sun, Jul 11, 2021 at 11:10 PM Jens Maurer via SG16 <sg16_at_[hidden] <mailto:sg16_at_[hidden]>> wrote:

> Regarding the uses of "determination"; this word (despite the
> considerable effort on clarifications around it) still feels
> that the compiler determines something (possibly by magic), but
> in reality it is the user that conveys an (out-of-band) assertion
> on the encoding of the source file. It feels English ought to
> have a word or phrase that fits better than "determination".
>
>
> We do not want to force a user input here, even if we hope that compilers will provide a sensible interface for users to use.
> The compiler can have any heuristic it desires, from "i know it's utf-8 because all files on this system are utf-8" to "reading the tea leaves"
> and I thought there was agreement on that now that we have wording to make bom irrelevant.
>
> I think ascertain is too strong, everything else I can think of is similar in meaning to determines.
> Decides leaves more room for error (which is what we want), but I don't know if "decides" applies to compilers.
> "Assumed in an implementation-defined manner" is more accurate, possibly. but awful.

We still have a strong preference that the compiler rejects ill-formed
UTF-8 files (i.e. those files that appear to be UTF-8, but are not properly
encoded), and we want to discourage the pretext of "uh, after fully reading
the file, I've determined it wasn't UTF-8 after all, so I used the
implementation-defined mapping here".

We may have insufficient scope in the standard to fully avoid that
situation, but I still feel that "determines" has the wrong connotation
here.

Jens