On 5/1/25 11:21 AM, Alisdair Meredith wrote:

Thanks Tom.

I will admit to mild confusion on the rationale for keeping the code conversion facets,

unless we also want to withdraw support for adopting https://wg21.link/P2783 for C++26

as well?

I don't see any reason to change direction for https://wg21.link/P2783 for C++26. The facets removed by that paper have been deprecated for multiple standardization cycles. The new concern with removing the charN_t <-> char8_t facets is that we haven't had a deprecation cycle and that we (unexpectedly) found some uses in the wild.

Could I request guidance on what a replacement facility would look like, given the

current facility is built to support a framework based on runtime locales?

There are several relevant proposals.

WG14 N3366: Restartable Functions for Efficient Character Conversions
P1629: Standard Text Encoding
P2728: Unicode in the Library, Part 1: UTF Transcoding

I understand that we do not want to remove something without a replacement, but

we will need a whole new text framework where character encoding is independent

of locale that do not have today. Mixing Unicode with runtime locales is a recipe for

text corruption, with no viable error handling in our current framework.

I agree. My perception of the consensus is that we favor removal (with or without a replacement) in this case so long as there is a deprecation cycle. We'll, of course, re-poll removal again pending a paper for C++29 or later that proposes removal. I imagine you'll provide us such an opportunity 😉

Tom.

AlisdairM

On Apr 23, 2025, at 11:22 PM, Tom Honermann <tom@honermann.net> wrote:

Thank you to everyone that attended today.

WG21 members can find the draft meeting summary with poll results at https://wiki.edg.com/bin/view/Wg21telecons2025/SG16Teleconference2025-04-23.

Tom.