Except that its numbers are underspecified and cannot be used safely outside of a certain range. The spec explicitly states that the precision of numbers is not

Except that its numbers are underspecified and cannot be used safely outside of a certain range. The spec explicitly states that the precision of numbers is not defined, meaning that N and N+1 may be the same number, and its behaviour would depend on the parser you're using. The number one rule when creating a serialisation format should be that serialisation and deserialisation is predictable. It's quite remarkable that two of the most popular formats doesn't do this. I'm actually surprised we haven't seen any major security issues caused by this.