My apologies. I did not actually intend to commit the directory.register. It was there for testing and I just did not see it. I don't really know to resubmit a patch starting from where I am now, but I will puzzle it out somehow. Is the .img file acceptable?
About "class scope": it really is an official part of Python: I did not make it up. It's exactly what we wanted to use here: a variable that is only accessible within a class, but not in its superclasses, but is a single variable shared amongst all members of the class. I am new to Python, so I do not know which language constructs are conventionally simply not used by developers. I will change this to conform to norms. Note that many of the scalars (e.g., MAX_MEM_SLOT) are a class scope, not instance scope.
I found out about class scope during testing, when working with group_descriptions. I thought I could simply modify the global from inside the FT-65 class. Nope, that affected the FT4. Then I figured I could just append in FT-65 __init__. Nope, that causes the append to occur multiple times if the main code instantiates multiple FT-65s. That's when I finally figured out the differences between global, instance, and class variables in Python. If the driver were only ever used in code that only works with a single radio during any particular run, I would never have seen this.