> While we're on the subject, though, trying to convert a very large int to float fails with an OverflowError. Perhaps Decimal should do the same?

I've always viewed float() as a cast. For casting Decimal's behavior
seems to be the right one to me.

If we go ahead with implicit casts for isclose(), I still think
the PEP's non-float section should be changed:  It sounds as if
native support was planned for Decimal.

Does someone have the tracker id of Chris?