On Thursday, 2 April 2026 02:45:34 Pacific Daylight Time Muneem via Std-
Proposals wrote:
> That's the point, std variant isn't usable here, nor is any other
> polymorphic technique, hence we need the features that I described in the
> standard.

You have not proven that. You have a hypothesis and a single benchmark that
seems to support your case (but falls apart under inspection). You need to
either argue that:
 a) it's an inherent problem that cannot be made better, however smart the
compiler can reasonably be; or
 b) a sufficiently smart compiler could do it, but it's too complex to wait for
it, so we should provide a fix for this large problem in a different way.

For option (a), you need to prove by arguing from the architecture, not a
benchmark. A proof requires showing that the information will never reasonably
be there.

For option (b), you need to show the problem is sufficiently great that it
warrants this solution. You have not done so.

And please learn to use emails. This is not a chat. Take the time to write
your emails and think things through. It's ok to reply to oneself occasionally
because you've just thought of something after pressing Send, but that should
be the exception, not the rule. You replied to my email 4 times in 20 minutes.
Please realise the impression you're giving when doing that: your emails are
not thought through, and you're just dumping at us the first thing that comes
to mind. Take the time, we're not going anywhere. If you want to see this
change through, it will take a minimum of 3 years anyway.

-- 
Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
  Principal Engineer - Intel Data Center - Platform & Sys. Eng.