It makes sense if we have signed zeros; then +0 represents an infinitesimal positive value. IEEE 754 has signed zeros, though the signs may sometimes not match programmers' expectations, e.g. 1 - 1e-50 - 1 == +0.
Maybe 1/+0 == +inf is reasonable, but would you agree that there's no reasonable answer for 0/0? So IMO, reasonable languages must define division as a partial function, or a function that can throw an error (or return NaN, Nothing, etc).
And if division is a partial function anyway, it seems best to not allow division by zero at all, since even 1/+0 == +inf can have unexpected consequences.
It makes sense if we have signed zeros; then +0 represents an infinitesimal positive value. IEEE 754 has signed zeros, though the signs may sometimes not match programmers' expectations, e.g. 1 - 1e-50 - 1 == +0.
Maybe 1/+0 == +inf is reasonable, but would you agree that there's no reasonable answer for 0/0? So IMO, reasonable languages must define division as a partial function, or a function that can throw an error (or return NaN, Nothing, etc).
And if division is a partial function anyway, it seems best to not allow division by zero at all, since even 1/+0 == +inf can have unexpected consequences.