I think with those changes, my disagreement would become a mere quibble. > I would still advocate for large signed types over unsigned types for most domain-le

I think with those changes, my disagreement would become a mere quibble. > I would still advocate for large signed types over unsigned types for most domain-level measurements. Even if you think you "can't" have a negative balance or distance field, use a signed integer type so that underflows are more correct. I agree with this, but I think I would personally still use unsigned types simulating two's complement that gives the correct underflow semantics. Yeah, I'm a hard egg.