Blog Post: Beating std::visit Without Really Trying

[Joseph Rushton Wakeling]

On Sunday, 6 October 2019 at 08:27:35 UTC, Seb wrote:
> Well, my guess it will be similar [...]

If you're not the one making those decisions it may be better not 
to prejudge them. A significant performance improvement is a 
different beast to moderate API/usability improvements.

> A standard library is _not_ supposed to be a place where actual 
> battle-testing happens. It's where things move when they have 
> been tested and no longer change.

You misunderstand what I mean by "battle-testing". Clearly 
designs should go through a high level of testing and usage 
before they go anywhere near the standard library. But the very 
fact of being placed in the standard library exposes them to 
orders of magnitude more usage, and hence gives a much stronger 
guarantee of establishing their correctness (or identifying their 
flaws).

It's much better to get newer and apparently better designs 
exposed to this level of scrutiny _before_ making them the only 
option in a new major release. That way you are much less likely 
to get hit by a showstopper edge case that no one anticipated.