Message11446

Julien, sounds good about your workaround. In general, when we find a discrepancy between CPython and Jython, we converge on CPython, even if it's arbitrary and undocumented/untested as is the case here ("" vs None/null). The reason we do so is that the larger ecosystem of Python code that's out there often relies on these distinctions. The Java API then usually just follows for simplicity's sake.

Because this will have an impact on users like you, we definitely will want to involve you in testing so that you can update accordingly. Yes, there's a bit of cost here - accepting either "" or null - but it's unavoidable IMHO.