Why is size_t unsigned?

[John Colvin]

On Monday, 22 July 2013 at 11:56:35 UTC, Regan Heath wrote:
> If we were to design the perfect type for representing a size 
> or length it would hold the maximum value of an unsigned int, 
> but would not undeflow to max unsigned int, instead it would 
> truncate.
>
> This type would have to be built on top of the existing 
> primitives and would therefore be less performant, which is a 
> shame and likely the reason it doesn't already exist.
>
> R

ARM directly supports saturating arithmetic, so this approach 
could be both practical and fast. However, x86 only supports it 
in SIMD, so it's not ideal.