On Wednesday, 6 November 2019 14:37:59 PST Brian Bi wrote:

The version of the story I heard is that if there's no type that can
represent all values of both types being compared, then the comparison
still has to yield the correct result, even though it might be "slow".

I don't think "slow" is acceptable. There's WAY too much code that performs 
comparisons between signed and unsigned that would be affected.

Then how about "error out"?

Shachar