I don't agree with the preference expressed in the guide, but I think I understand the issue. Consider this Ruby code:

class A 
  class << self 
    attr_accessor :some_attribute 
  end 
end 

class B < A 
end 

As far as I know, that's the only way to support usage like this:

A.some_attribute = :value_1 
B.some_attribute # => nil 
B.some_attribute = :value_2 
A.some_attribute # => :value_1 

I mean, using a class variable would cause it to be shared by A and B, which we don't want.

Does that sound right to you? Is there another way to implement that snippet above?

I think the same thing is true for alias, it only works for class methods inside class << self.

So, I don't agree with the preference, but I think that explains the difference, does that make sense?

Originally posted by @rmosolgo in #7 (comment)