In Python 2.x and 3.2, I used to use a Request subclass I created for overriding the method used:class MethodRequest(request.Request):def __init__(self, *args, **kwargs):"""Construct a MethodRequest. Usage is the same as for`urllib.request.Request` except it also takes an optional `method`keyword argument. If supplied, `method` will be used instead ofthe default."""if 'method' in kwargs:self.method = kwargs.pop('method')return request.Request.__init__(self, *args, **kwargs)def get_method(self):return getattr(self, 'method', request.Request.get_method(self))In Python 3.3, which now supports a method parameter, it broke this paradigm (because the method is stored in the instance and is always set to None in __init__ if not specified).I believe a paradigm where the method is stored as a class attribute and possibly overridden in an instance would be much better, allowing for subclasses to simply and directly override the method. For example:class HeadRequest(MethodRequest):method = 'HEAD'That straightforward example works very well if method is allowed to be a class attribute, but won't work at all if 'method' is always set as an instance attribute in __init__.And while it's possible for HeadRequest to override __init__, that requires HeadRequest to override that entire signature, which is less elegant than simply setting a class attribute.For Python 3.4, I'd like to adapt the Request class to allow the Method to be defined at the class level (while still honoring customization at the instance level).