What real status of this document? As I figured out py3k trunk uses this ones. Most part of 'battery included' modules don't use this feature, leaving m_traverse, m_clear and m_free NULL pointers. There are only exception: _io. But looks like all underground python machinery is already ported to support PEP. Maybe it's good point to update PEP? "accepted; may not be implemented yet" sounds a bit confusing. Just insert somethere close to top of content: "Implemented by Python itself, but builtin modules should be converted". Next question: are where all 'python standard library' modules looks like "not converted yet" really need to be refactored? Or we can split this huge set to 'looks already good' and 'requires some work'? And just add comment for every PyModuleDef definition like: "Completely satisfy PEP 3121, signed and stamped by Martin von Loewis and GvR".? Last question: PEP refers to "multiple interpreters" feature of python. This feature presented at least in 2.xx branch (don't remember about 1.xx - up to 2003 I used python as scripting language only and not care about C interface). I never saw usage of this one in real use cases. Is there projects (preferable with code sources, of course) what runs multiple interpreter in same process?