On Friday, 21 March 2025 03:47:22 Pacific Daylight Time Hans Åberg via Std-
Proposals wrote:
> It is the link time choices I have mind, possibly requiring the linker to be
> smarter.

The current solutions are about the application code overriding the standard
implementation, not the other way around. The existing linker solutions will
not work for what you're proposing.

Requiring linker changes imposes an additional time constraint for a lot of
platforms, because often the linker comes from the OS vendor (who won't update
their content) while the compiler comes from a third-party vendor. You'll need
a very, very good use-case for this and have to accept that it won't be a
commonplace occurrence for at least another 10 years. In turn, that means the
number of Standard Library features that could use the feature is smaller.

I have not yet heard a single use-case for this. Providing a double-underscore
function from the runtime is not a valid use-case, nor does it *work* as
described anyway.

-- 
Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
  Principal Engineer - Intel DCAI Platform & System Engineering