In the section on reducing floating point numbers to integers, I believe it would be appropriate to clearly specify behavior of other floating point reductions, e.g. reducing 0.5 from double to a half-width float.
In addition, I believe the behavior needs to be defined as to what an application should do when conversion of an integer to a floating point number is in itself lossy, e.g. what to do if an application expects a double but the CBOR value received is the decimal 9007199254740993. Silent truncation would make the round-trip non-deterministic.
In the section on reducing floating point numbers to integers, I believe it would be appropriate to clearly specify behavior of other floating point reductions, e.g. reducing 0.5 from double to a half-width float.
In addition, I believe the behavior needs to be defined as to what an application should do when conversion of an integer to a floating point number is in itself lossy, e.g. what to do if an application expects a double but the CBOR value received is the decimal 9007199254740993. Silent truncation would make the round-trip non-deterministic.