Improve the OOP ABI

[Walter Bright]

On 1/22/2024 2:49 PM, Richard (Rikki) Andrew Cattermole wrote:
>> One collision in 6 billion hashes per second for 100 years.
> 
> If we were gambling on something not happening, those are great odds.
> 
> However if this does occur, silent memory corruption is possible.
> 
> I don't want to be gambling on whether there will be silent memory corruption.
> 
> This is my perspective on the topic anyway. I love guarantees of program security.

I do understand the desire for mathematical perfection, and feel it myself. But 
it has a significant and always present performance penalty. Engineering is 
always about tradeoffs. There's no getting away from it.

I'm not concerned about a meteor hitting my house, either, which is far far more 
likely.

BTW, D uses md5 hashes to enable the linker to merge redundant string literals.

Other compromises we make for performance include things like floating point 
rounding.