All three (Python, Java, .NET):
- _CRITICAL FIX:_ Encoding/serializing unicode strings is now done by `repr` itself instead of custom code. This solves the issue that previously invalid serialized data which was unparsable, could have been generated from unicode strings.
- Added Serializer.MaximumLevel to avoid too deep recursion resulting in stack overflow errors.
- Serializer.Maximumlevel set to a reasonable value (500, or less if your python's recursionlimit is small)
Python:
- Improved Jython compatibility (built-in support for several Jython types that were previously unserializable such as PyByteArray).
- Better error when invalid serialized data is used (containing 0-bytes for instance)
- Improved serialization performance.