I have implemented CBOR for Python and Go. In my testing: CBOR+GZIP is slightly smaller, about 10%, than JSON+GZIP CBOR also parses faster, sometimes as much as

I have implemented CBOR for Python and Go. In my testing: CBOR+GZIP is slightly smaller, about 10%, than JSON+GZIP CBOR also parses faster, sometimes as much as 3-5x speedup. I'm not sure what 'ambiguous decoding' you're referring to. I've seen some complains that an int might decode to an int32 or int64 depending on the value but be different than the original storage size. e.g. '42' stored from an int64 might unpack to an int8 depending on the language and implementation of the decoder.